Hugging Face's logo

baa-ai
/
Merino-Mini

Sentence Similarity
sentence-transformers
Safetensors
baa-embedding-reranker
retrieval
embeddings
reranker
cross-encoder
rag
Model card Files
xet
 Community
tomkay commited on
Commit
8e640bf
·
verified ·
1 Parent(s): 707db43

Add baa.ai Embedding-Reranker-BGE-Small-L6-v1 (unified shared word-emb)

Browse files
Files changed (18) hide show
  1. LICENSE +202 -0
  2. NOTICE +27 -0
  3. README.md +64 -0
  4. config.json +23 -0
  5. embedder/1_Pooling/config.json +5 -0
  6. embedder/README.md +3074 -0
  7. embedder/config.json +35 -0
  8. embedder/config_sentence_transformers.json +14 -0
  9. embedder/model.safetensors +3 -0
  10. embedder/modules.json +20 -0
  11. embedder/sentence_bert_config.json +10 -0
  12. embedder/tokenizer.json +0 -0
  13. embedder/tokenizer_config.json +18 -0
  14. modeling_baa.py +83 -0
  15. reranker/config.json +37 -0
  16. reranker/model.safetensors +3 -0
  17. reranker/tokenizer.json +0 -0
  18. reranker/tokenizer_config.json +18 -0
LICENSE ADDED
@@ -0,0 +1,202 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+
2
+ Apache License
3
+ Version 2.0, January 2004
4
+ http://www.apache.org/licenses/
5
+
6
+ TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
7
+
8
+ 1. Definitions.
9
+
10
+ "License" shall mean the terms and conditions for use, reproduction,
11
+ and distribution as defined by Sections 1 through 9 of this document.
12
+
13
+ "Licensor" shall mean the copyright owner or entity authorized by
14
+ the copyright owner that is granting the License.
15
+
16
+ "Legal Entity" shall mean the union of the acting entity and all
17
+ other entities that control, are controlled by, or are under common
18
+ control with that entity. For the purposes of this definition,
19
+ "control" means (i) the power, direct or indirect, to cause the
20
+ direction or management of such entity, whether by contract or
21
+ otherwise, or (ii) ownership of fifty percent (50%) or more of the
22
+ outstanding shares, or (iii) beneficial ownership of such entity.
23
+
24
+ "You" (or "Your") shall mean an individual or Legal Entity
25
+ exercising permissions granted by this License.
26
+
27
+ "Source" form shall mean the preferred form for making modifications,
28
+ including but not limited to software source code, documentation
29
+ source, and configuration files.
30
+
31
+ "Object" form shall mean any form resulting from mechanical
32
+ transformation or translation of a Source form, including but
33
+ not limited to compiled object code, generated documentation,
34
+ and conversions to other media types.
35
+
36
+ "Work" shall mean the work of authorship, whether in Source or
37
+ Object form, made available under the License, as indicated by a
38
+ copyright notice that is included in or attached to the work
39
+ (an example is provided in the Appendix below).
40
+
41
+ "Derivative Works" shall mean any work, whether in Source or Object
42
+ form, that is based on (or derived from) the Work and for which the
43
+ editorial revisions, annotations, elaborations, or other modifications
44
+ represent, as a whole, an original work of authorship. For the purposes
45
+ of this License, Derivative Works shall not include works that remain
46
+ separable from, or merely link (or bind by name) to the interfaces of,
47
+ the Work and Derivative Works thereof.
48
+
49
+ "Contribution" shall mean any work of authorship, including
50
+ the original version of the Work and any modifications or additions
51
+ to that Work or Derivative Works thereof, that is intentionally
52
+ submitted to Licensor for inclusion in the Work by the copyright owner
53
+ or by an individual or Legal Entity authorized to submit on behalf of
54
+ the copyright owner. For the purposes of this definition, "submitted"
55
+ means any form of electronic, verbal, or written communication sent
56
+ to the Licensor or its representatives, including but not limited to
57
+ communication on electronic mailing lists, source code control systems,
58
+ and issue tracking systems that are managed by, or on behalf of, the
59
+ Licensor for the purpose of discussing and improving the Work, but
60
+ excluding communication that is conspicuously marked or otherwise
61
+ designated in writing by the copyright owner as "Not a Contribution."
62
+
63
+ "Contributor" shall mean Licensor and any individual or Legal Entity
64
+ on behalf of whom a Contribution has been received by Licensor and
65
+ subsequently incorporated within the Work.
66
+
67
+ 2. Grant of Copyright License. Subject to the terms and conditions of
68
+ this License, each Contributor hereby grants to You a perpetual,
69
+ worldwide, non-exclusive, no-charge, royalty-free, irrevocable
70
+ copyright license to reproduce, prepare Derivative Works of,
71
+ publicly display, publicly perform, sublicense, and distribute the
72
+ Work and such Derivative Works in Source or Object form.
73
+
74
+ 3. Grant of Patent License. Subject to the terms and conditions of
75
+ this License, each Contributor hereby grants to You a perpetual,
76
+ worldwide, non-exclusive, no-charge, royalty-free, irrevocable
77
+ (except as stated in this section) patent license to make, have made,
78
+ use, offer to sell, sell, import, and otherwise transfer the Work,
79
+ where such license applies only to those patent claims licensable
80
+ by such Contributor that are necessarily infringed by their
81
+ Contribution(s) alone or by combination of their Contribution(s)
82
+ with the Work to which such Contribution(s) was submitted. If You
83
+ institute patent litigation against any entity (including a
84
+ cross-claim or counterclaim in a lawsuit) alleging that the Work
85
+ or a Contribution incorporated within the Work constitutes direct
86
+ or contributory patent infringement, then any patent licenses
87
+ granted to You under this License for that Work shall terminate
88
+ as of the date such litigation is filed.
89
+
90
+ 4. Redistribution. You may reproduce and distribute copies of the
91
+ Work or Derivative Works thereof in any medium, with or without
92
+ modifications, and in Source or Object form, provided that You
93
+ meet the following conditions:
94
+
95
+ (a) You must give any other recipients of the Work or
96
+ Derivative Works a copy of this License; and
97
+
98
+ (b) You must cause any modified files to carry prominent notices
99
+ stating that You changed the files; and
100
+
101
+ (c) You must retain, in the Source form of any Derivative Works
102
+ that You distribute, all copyright, patent, trademark, and
103
+ attribution notices from the Source form of the Work,
104
+ excluding those notices that do not pertain to any part of
105
+ the Derivative Works; and
106
+
107
+ (d) If the Work includes a "NOTICE" text file as part of its
108
+ distribution, then any Derivative Works that You distribute must
109
+ include a readable copy of the attribution notices contained
110
+ within such NOTICE file, excluding those notices that do not
111
+ pertain to any part of the Derivative Works, in at least one
112
+ of the following places: within a NOTICE text file distributed
113
+ as part of the Derivative Works; within the Source form or
114
+ documentation, if provided along with the Derivative Works; or,
115
+ within a display generated by the Derivative Works, if and
116
+ wherever such third-party notices normally appear. The contents
117
+ of the NOTICE file are for informational purposes only and
118
+ do not modify the License. You may add Your own attribution
119
+ notices within Derivative Works that You distribute, alongside
120
+ or as an addendum to the NOTICE text from the Work, provided
121
+ that such additional attribution notices cannot be construed
122
+ as modifying the License.
123
+
124
+ You may add Your own copyright statement to Your modifications and
125
+ may provide additional or different license terms and conditions
126
+ for use, reproduction, or distribution of Your modifications, or
127
+ for any such Derivative Works as a whole, provided Your use,
128
+ reproduction, and distribution of the Work otherwise complies with
129
+ the conditions stated in this License.
130
+
131
+ 5. Submission of Contributions. Unless You explicitly state otherwise,
132
+ any Contribution intentionally submitted for inclusion in the Work
133
+ by You to the Licensor shall be under the terms and conditions of
134
+ this License, without any additional terms or conditions.
135
+ Notwithstanding the above, nothing herein shall supersede or modify
136
+ the terms of any separate license agreement you may have executed
137
+ with Licensor regarding such Contributions.
138
+
139
+ 6. Trademarks. This License does not grant permission to use the trade
140
+ names, trademarks, service marks, or product names of the Licensor,
141
+ except as required for reasonable and customary use in describing the
142
+ origin of the Work and reproducing the content of the NOTICE file.
143
+
144
+ 7. Disclaimer of Warranty. Unless required by applicable law or
145
+ agreed to in writing, Licensor provides the Work (and each
146
+ Contributor provides its Contributions) on an "AS IS" BASIS,
147
+ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
148
+ implied, including, without limitation, any warranties or conditions
149
+ of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
150
+ PARTICULAR PURPOSE. You are solely responsible for determining the
151
+ appropriateness of using or redistributing the Work and assume any
152
+ risks associated with Your exercise of permissions under this License.
153
+
154
+ 8. Limitation of Liability. In no event and under no legal theory,
155
+ whether in tort (including negligence), contract, or otherwise,
156
+ unless required by applicable law (such as deliberate and grossly
157
+ negligent acts) or agreed to in writing, shall any Contributor be
158
+ liable to You for damages, including any direct, indirect, special,
159
+ incidental, or consequential damages of any character arising as a
160
+ result of this License or out of the use or inability to use the
161
+ Work (including but not limited to damages for loss of goodwill,
162
+ work stoppage, computer failure or malfunction, or any and all
163
+ other commercial damages or losses), even if such Contributor
164
+ has been advised of the possibility of such damages.
165
+
166
+ 9. Accepting Warranty or Additional Liability. While redistributing
167
+ the Work or Derivative Works thereof, You may choose to offer,
168
+ and charge a fee for, acceptance of support, warranty, indemnity,
169
+ or other liability obligations and/or rights consistent with this
170
+ License. However, in accepting such obligations, You may act only
171
+ on Your own behalf and on Your sole responsibility, not on behalf
172
+ of any other Contributor, and only if You agree to indemnify,
173
+ defend, and hold each Contributor harmless for any liability
174
+ incurred by, or claims asserted against, such Contributor by reason
175
+ of your accepting any such warranty or additional liability.
176
+
177
+ END OF TERMS AND CONDITIONS
178
+
179
+ APPENDIX: How to apply the Apache License to your work.
180
+
181
+ To apply the Apache License to your work, attach the following
182
+ boilerplate notice, with the fields enclosed by brackets "[]"
183
+ replaced with your own identifying information. (Don't include
184
+ the brackets!) The text should be enclosed in the appropriate
185
+ comment syntax for the file format. We also recommend that a
186
+ file or class name and description of purpose be included on the
187
+ same "printed page" as the copyright notice for easier
188
+ identification within third-party archives.
189
+
190
+ Copyright [yyyy] [name of copyright owner]
191
+
192
+ Licensed under the Apache License, Version 2.0 (the "License");
193
+ you may not use this file except in compliance with the License.
194
+ You may obtain a copy of the License at
195
+
196
+ http://www.apache.org/licenses/LICENSE-2.0
197
+
198
+ Unless required by applicable law or agreed to in writing, software
199
+ distributed under the License is distributed on an "AS IS" BASIS,
200
+ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
201
+ See the License for the specific language governing permissions and
202
+ limitations under the License.
NOTICE ADDED
@@ -0,0 +1,27 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ baa.ai · Embedding-Reranker-BGE-Small-L6-v1
2
+ Copyright (c) 2026 baa.ai
3
+
4
+ This product is a derivative work licensed under the Apache License, Version 2.0.
5
+ It re-packages and modifies the following upstream models:
6
+
7
+ 1. BAAI/bge-small-en-v1.5 (license: MIT)
8
+ Used as the bi-encoder embedder. Provides the canonical (shared)
9
+ word-embedding table for the combined model.
10
+ https://huggingface.co/BAAI/bge-small-en-v1.5
11
+
12
+ 2. cross-encoder/ms-marco-MiniLM-L-6-v2 (license: Apache-2.0)
13
+ Used as the cross-encoder reranker. Its word-embedding table has been
14
+ removed on disk and is injected at load time from the shared table
15
+ above, reducing the combined footprint.
16
+ https://huggingface.co/cross-encoder/ms-marco-MiniLM-L-6-v2
17
+
18
+ Both upstream models derive from the BERT (uncased, 30522-vocab, 384-dim) architecture.
19
+
20
+ Modifications by baa.ai:
21
+ - Unified the two models into a single artifact over one shared
22
+ word-embedding table (the reranker's word-embedding matrix is stored
23
+ once, in the embedder, and injected at load).
24
+ - Added a combined loader (modeling_baa.py) exposing embed() and rerank().
25
+
26
+ This NOTICE file is provided in accordance with Section 4(d) of the
27
+ Apache License, Version 2.0. See the LICENSE file for the full license text.
README.md ADDED
@@ -0,0 +1,64 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ library_name: sentence-transformers
4
+ tags:
5
+ - retrieval
6
+ - embeddings
7
+ - reranker
8
+ - cross-encoder
9
+ - rag
10
+ - sentence-similarity
11
+ pipeline_tag: sentence-similarity
12
+ ---
13
+
14
+ # baa.ai · Embedding-Reranker-BGE-Small-L6-v1
15
+
16
+ **A single model that does both halves of RAG retrieval — bi-encoder embedding *and* cross-encoder reranking — over one shared word-embedding table.**
17
+
18
+ The embedder and reranker share their word-embedding table (stored once), so the packaged model is **~20.9% smaller on disk than shipping the two components separately**, with **no measured loss in retrieval quality**.
19
+
20
+ ## Why this model
21
+
22
+ Most RAG stacks bolt an embedder onto a reranker and pay for both. A well-matched embedder + reranker that share a backbone recover the right documents just as well — in a single, smaller download.
23
+
24
+ - **Two jobs, one download** — embed for retrieval, then rerank the candidates.
25
+ - **Smaller footprint** — the shared word-embedding table is stored once.
26
+ - **Strong default** — a sensible starting point for production RAG.
27
+
28
+ ## Usage
29
+
30
+ ```python
31
+ from modeling_baa import BaaEmbeddingReranker # included in this repo
32
+
33
+ m = BaaEmbeddingReranker("baa-ai/Embedding-Reranker-BGE-Small-L6-v1")
34
+ qv = m.embed(["how does a cross-encoder reranker work?"], is_query=True)[0]
35
+ dv = m.embed(["a cross-encoder scores a (query, document) pair jointly",
36
+ "bi-encoders embed query and document separately for fast retrieval"])
37
+ ranked = m.rerank("how does a cross-encoder reranker work?",
38
+ ["a cross-encoder scores a (query, document) pair jointly",
39
+ "the mitochondria is the powerhouse of the cell"])
40
+ # -> [(doc, score), ...] sorted best-first
41
+ ```
42
+
43
+ ## Get the optimal models for *your* data
44
+
45
+ This model is a great **default**. But the best embedder and reranker are **corpus-specific**. **baa.ai offers exclusive tooling that identifies the optimal embedding and reranking models for your specific data** — if you want that tailored recommendation, **reach out to baa.ai**.
46
+
47
+ ## Specs
48
+
49
+ | | |
50
+ |---|---|
51
+ | Embedding dim | 384 |
52
+ | Vocab | 30522 |
53
+ | Max sequence length | 512 |
54
+ | Combined params | ~44.4M (vs ~56.1M separate) |
55
+ | Footprint vs separate models | ~20.9% smaller on disk, no measured quality loss |
56
+ | Hardware | CPU / edge / GPU |
57
+
58
+ ## License & attribution
59
+
60
+ Released under the **Apache License 2.0**. Derivative work re-packaging two upstream models into a single shared-backbone artifact (see `NOTICE`):
61
+ - `BAAI/bge-small-en-v1.5` (embedder)
62
+ - `cross-encoder/ms-marco-MiniLM-L-6-v2` (reranker)
63
+
64
+ © baa.ai. Provided "as is" without warranty; see `LICENSE`.
config.json ADDED
@@ -0,0 +1,23 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model_type": "baa-embed-rerank",
3
+ "description": "Unified bi-encoder embedder + cross-encoder reranker sharing one word-embedding table.",
4
+ "components": {
5
+ "embedder": "embedder/",
6
+ "reranker": "reranker/"
7
+ },
8
+ "loader": "modeling_baa.BaaEmbeddingReranker",
9
+ "embed_query_prompt": "Represent this sentence for searching relevant passages: ",
10
+ "embed_doc_prompt": "",
11
+ "embedding_dim": 384,
12
+ "vocab_size": 30522,
13
+ "max_seq_length": 512,
14
+ "separate_params_millions": 56.1,
15
+ "combined_params_millions": 44.4,
16
+ "disk_saving_pct": 20.9,
17
+ "license": "apache-2.0",
18
+ "trust_remote_code": false,
19
+ "upstream": {
20
+ "embedder": "BAAI/bge-small-en-v1.5",
21
+ "reranker": "cross-encoder/ms-marco-MiniLM-L-6-v2"
22
+ }
23
+ }
embedder/1_Pooling/config.json ADDED
@@ -0,0 +1,5 @@
 
 
 
 
 
 
1
+ {
2
+ "embedding_dimension": 384,
3
+ "pooling_mode": "cls",
4
+ "include_prompt": true
5
+ }
embedder/README.md ADDED
@@ -0,0 +1,3074 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ tags:
3
+ - sentence-transformers
4
+ - feature-extraction
5
+ - sentence-similarity
6
+ - transformers
7
+ - mteb
8
+ model-index:
9
+ - name: bge-small-en-v1.5
10
+ results:
11
+ - task:
12
+ type: Classification
13
+ dataset:
14
+ type: mteb/amazon_counterfactual
15
+ name: MTEB AmazonCounterfactualClassification (en)
16
+ config: en
17
+ split: test
18
+ revision: e8379541af4e31359cca9fbcf4b00f2671dba205
19
+ metrics:
20
+ - type: accuracy
21
+ value: 73.79104477611939
22
+ - type: ap
23
+ value: 37.21923821573361
24
+ - type: f1
25
+ value: 68.0914945617093
26
+ - task:
27
+ type: Classification
28
+ dataset:
29
+ type: mteb/amazon_polarity
30
+ name: MTEB AmazonPolarityClassification
31
+ config: default
32
+ split: test
33
+ revision: e2d317d38cd51312af73b3d32a06d1a08b442046
34
+ metrics:
35
+ - type: accuracy
36
+ value: 92.75377499999999
37
+ - type: ap
38
+ value: 89.46766124546022
39
+ - type: f1
40
+ value: 92.73884001331487
41
+ - task:
42
+ type: Classification
43
+ dataset:
44
+ type: mteb/amazon_reviews_multi
45
+ name: MTEB AmazonReviewsClassification (en)
46
+ config: en
47
+ split: test
48
+ revision: 1399c76144fd37290681b995c656ef9b2e06e26d
49
+ metrics:
50
+ - type: accuracy
51
+ value: 46.986
52
+ - type: f1
53
+ value: 46.55936786727896
54
+ - task:
55
+ type: Retrieval
56
+ dataset:
57
+ type: arguana
58
+ name: MTEB ArguAna
59
+ config: default
60
+ split: test
61
+ revision: None
62
+ metrics:
63
+ - type: map_at_1
64
+ value: 35.846000000000004
65
+ - type: map_at_10
66
+ value: 51.388
67
+ - type: map_at_100
68
+ value: 52.132999999999996
69
+ - type: map_at_1000
70
+ value: 52.141000000000005
71
+ - type: map_at_3
72
+ value: 47.037
73
+ - type: map_at_5
74
+ value: 49.579
75
+ - type: mrr_at_1
76
+ value: 36.558
77
+ - type: mrr_at_10
78
+ value: 51.658
79
+ - type: mrr_at_100
80
+ value: 52.402
81
+ - type: mrr_at_1000
82
+ value: 52.410000000000004
83
+ - type: mrr_at_3
84
+ value: 47.345
85
+ - type: mrr_at_5
86
+ value: 49.797999999999995
87
+ - type: ndcg_at_1
88
+ value: 35.846000000000004
89
+ - type: ndcg_at_10
90
+ value: 59.550000000000004
91
+ - type: ndcg_at_100
92
+ value: 62.596
93
+ - type: ndcg_at_1000
94
+ value: 62.759
95
+ - type: ndcg_at_3
96
+ value: 50.666999999999994
97
+ - type: ndcg_at_5
98
+ value: 55.228
99
+ - type: precision_at_1
100
+ value: 35.846000000000004
101
+ - type: precision_at_10
102
+ value: 8.542
103
+ - type: precision_at_100
104
+ value: 0.984
105
+ - type: precision_at_1000
106
+ value: 0.1
107
+ - type: precision_at_3
108
+ value: 20.389
109
+ - type: precision_at_5
110
+ value: 14.438
111
+ - type: recall_at_1
112
+ value: 35.846000000000004
113
+ - type: recall_at_10
114
+ value: 85.42
115
+ - type: recall_at_100
116
+ value: 98.43499999999999
117
+ - type: recall_at_1000
118
+ value: 99.644
119
+ - type: recall_at_3
120
+ value: 61.166
121
+ - type: recall_at_5
122
+ value: 72.191
123
+ - task:
124
+ type: Clustering
125
+ dataset:
126
+ type: mteb/arxiv-clustering-p2p
127
+ name: MTEB ArxivClusteringP2P
128
+ config: default
129
+ split: test
130
+ revision: a122ad7f3f0291bf49cc6f4d32aa80929df69d5d
131
+ metrics:
132
+ - type: v_measure
133
+ value: 47.402770198163594
134
+ - task:
135
+ type: Clustering
136
+ dataset:
137
+ type: mteb/arxiv-clustering-s2s
138
+ name: MTEB ArxivClusteringS2S
139
+ config: default
140
+ split: test
141
+ revision: f910caf1a6075f7329cdf8c1a6135696f37dbd53
142
+ metrics:
143
+ - type: v_measure
144
+ value: 40.01545436974177
145
+ - task:
146
+ type: Reranking
147
+ dataset:
148
+ type: mteb/askubuntudupquestions-reranking
149
+ name: MTEB AskUbuntuDupQuestions
150
+ config: default
151
+ split: test
152
+ revision: 2000358ca161889fa9c082cb41daa8dcfb161a54
153
+ metrics:
154
+ - type: map
155
+ value: 62.586465273207196
156
+ - type: mrr
157
+ value: 74.42169019038825
158
+ - task:
159
+ type: STS
160
+ dataset:
161
+ type: mteb/biosses-sts
162
+ name: MTEB BIOSSES
163
+ config: default
164
+ split: test
165
+ revision: d3fb88f8f02e40887cd149695127462bbcf29b4a
166
+ metrics:
167
+ - type: cos_sim_pearson
168
+ value: 85.1891186537969
169
+ - type: cos_sim_spearman
170
+ value: 83.75492046087288
171
+ - type: euclidean_pearson
172
+ value: 84.11766204805357
173
+ - type: euclidean_spearman
174
+ value: 84.01456493126516
175
+ - type: manhattan_pearson
176
+ value: 84.2132950502772
177
+ - type: manhattan_spearman
178
+ value: 83.89227298813377
179
+ - task:
180
+ type: Classification
181
+ dataset:
182
+ type: mteb/banking77
183
+ name: MTEB Banking77Classification
184
+ config: default
185
+ split: test
186
+ revision: 0fd18e25b25c072e09e0d92ab615fda904d66300
187
+ metrics:
188
+ - type: accuracy
189
+ value: 85.74025974025975
190
+ - type: f1
191
+ value: 85.71493566466381
192
+ - task:
193
+ type: Clustering
194
+ dataset:
195
+ type: mteb/biorxiv-clustering-p2p
196
+ name: MTEB BiorxivClusteringP2P
197
+ config: default
198
+ split: test
199
+ revision: 65b79d1d13f80053f67aca9498d9402c2d9f1f40
200
+ metrics:
201
+ - type: v_measure
202
+ value: 38.467181385006434
203
+ - task:
204
+ type: Clustering
205
+ dataset:
206
+ type: mteb/biorxiv-clustering-s2s
207
+ name: MTEB BiorxivClusteringS2S
208
+ config: default
209
+ split: test
210
+ revision: 258694dd0231531bc1fd9de6ceb52a0853c6d908
211
+ metrics:
212
+ - type: v_measure
213
+ value: 34.719496037339056
214
+ - task:
215
+ type: Retrieval
216
+ dataset:
217
+ type: BeIR/cqadupstack
218
+ name: MTEB CQADupstackAndroidRetrieval
219
+ config: default
220
+ split: test
221
+ revision: None
222
+ metrics:
223
+ - type: map_at_1
224
+ value: 29.587000000000003
225
+ - type: map_at_10
226
+ value: 41.114
227
+ - type: map_at_100
228
+ value: 42.532
229
+ - type: map_at_1000
230
+ value: 42.661
231
+ - type: map_at_3
232
+ value: 37.483
233
+ - type: map_at_5
234
+ value: 39.652
235
+ - type: mrr_at_1
236
+ value: 36.338
237
+ - type: mrr_at_10
238
+ value: 46.763
239
+ - type: mrr_at_100
240
+ value: 47.393
241
+ - type: mrr_at_1000
242
+ value: 47.445
243
+ - type: mrr_at_3
244
+ value: 43.538
245
+ - type: mrr_at_5
246
+ value: 45.556000000000004
247
+ - type: ndcg_at_1
248
+ value: 36.338
249
+ - type: ndcg_at_10
250
+ value: 47.658
251
+ - type: ndcg_at_100
252
+ value: 52.824000000000005
253
+ - type: ndcg_at_1000
254
+ value: 54.913999999999994
255
+ - type: ndcg_at_3
256
+ value: 41.989
257
+ - type: ndcg_at_5
258
+ value: 44.944
259
+ - type: precision_at_1
260
+ value: 36.338
261
+ - type: precision_at_10
262
+ value: 9.156
263
+ - type: precision_at_100
264
+ value: 1.4789999999999999
265
+ - type: precision_at_1000
266
+ value: 0.196
267
+ - type: precision_at_3
268
+ value: 20.076
269
+ - type: precision_at_5
270
+ value: 14.85
271
+ - type: recall_at_1
272
+ value: 29.587000000000003
273
+ - type: recall_at_10
274
+ value: 60.746
275
+ - type: recall_at_100
276
+ value: 82.157
277
+ - type: recall_at_1000
278
+ value: 95.645
279
+ - type: recall_at_3
280
+ value: 44.821
281
+ - type: recall_at_5
282
+ value: 52.819
283
+ - task:
284
+ type: Retrieval
285
+ dataset:
286
+ type: BeIR/cqadupstack
287
+ name: MTEB CQADupstackEnglishRetrieval
288
+ config: default
289
+ split: test
290
+ revision: None
291
+ metrics:
292
+ - type: map_at_1
293
+ value: 30.239
294
+ - type: map_at_10
295
+ value: 39.989000000000004
296
+ - type: map_at_100
297
+ value: 41.196
298
+ - type: map_at_1000
299
+ value: 41.325
300
+ - type: map_at_3
301
+ value: 37.261
302
+ - type: map_at_5
303
+ value: 38.833
304
+ - type: mrr_at_1
305
+ value: 37.516
306
+ - type: mrr_at_10
307
+ value: 46.177
308
+ - type: mrr_at_100
309
+ value: 46.806
310
+ - type: mrr_at_1000
311
+ value: 46.849000000000004
312
+ - type: mrr_at_3
313
+ value: 44.002
314
+ - type: mrr_at_5
315
+ value: 45.34
316
+ - type: ndcg_at_1
317
+ value: 37.516
318
+ - type: ndcg_at_10
319
+ value: 45.586
320
+ - type: ndcg_at_100
321
+ value: 49.897000000000006
322
+ - type: ndcg_at_1000
323
+ value: 51.955
324
+ - type: ndcg_at_3
325
+ value: 41.684
326
+ - type: ndcg_at_5
327
+ value: 43.617
328
+ - type: precision_at_1
329
+ value: 37.516
330
+ - type: precision_at_10
331
+ value: 8.522
332
+ - type: precision_at_100
333
+ value: 1.374
334
+ - type: precision_at_1000
335
+ value: 0.184
336
+ - type: precision_at_3
337
+ value: 20.105999999999998
338
+ - type: precision_at_5
339
+ value: 14.152999999999999
340
+ - type: recall_at_1
341
+ value: 30.239
342
+ - type: recall_at_10
343
+ value: 55.03
344
+ - type: recall_at_100
345
+ value: 73.375
346
+ - type: recall_at_1000
347
+ value: 86.29599999999999
348
+ - type: recall_at_3
349
+ value: 43.269000000000005
350
+ - type: recall_at_5
351
+ value: 48.878
352
+ - task:
353
+ type: Retrieval
354
+ dataset:
355
+ type: BeIR/cqadupstack
356
+ name: MTEB CQADupstackGamingRetrieval
357
+ config: default
358
+ split: test
359
+ revision: None
360
+ metrics:
361
+ - type: map_at_1
362
+ value: 38.338
363
+ - type: map_at_10
364
+ value: 50.468999999999994
365
+ - type: map_at_100
366
+ value: 51.553000000000004
367
+ - type: map_at_1000
368
+ value: 51.608
369
+ - type: map_at_3
370
+ value: 47.107
371
+ - type: map_at_5
372
+ value: 49.101
373
+ - type: mrr_at_1
374
+ value: 44.201
375
+ - type: mrr_at_10
376
+ value: 54.057
377
+ - type: mrr_at_100
378
+ value: 54.764
379
+ - type: mrr_at_1000
380
+ value: 54.791000000000004
381
+ - type: mrr_at_3
382
+ value: 51.56699999999999
383
+ - type: mrr_at_5
384
+ value: 53.05
385
+ - type: ndcg_at_1
386
+ value: 44.201
387
+ - type: ndcg_at_10
388
+ value: 56.379000000000005
389
+ - type: ndcg_at_100
390
+ value: 60.645
391
+ - type: ndcg_at_1000
392
+ value: 61.73499999999999
393
+ - type: ndcg_at_3
394
+ value: 50.726000000000006
395
+ - type: ndcg_at_5
396
+ value: 53.58500000000001
397
+ - type: precision_at_1
398
+ value: 44.201
399
+ - type: precision_at_10
400
+ value: 9.141
401
+ - type: precision_at_100
402
+ value: 1.216
403
+ - type: precision_at_1000
404
+ value: 0.135
405
+ - type: precision_at_3
406
+ value: 22.654
407
+ - type: precision_at_5
408
+ value: 15.723999999999998
409
+ - type: recall_at_1
410
+ value: 38.338
411
+ - type: recall_at_10
412
+ value: 70.30499999999999
413
+ - type: recall_at_100
414
+ value: 88.77199999999999
415
+ - type: recall_at_1000
416
+ value: 96.49799999999999
417
+ - type: recall_at_3
418
+ value: 55.218
419
+ - type: recall_at_5
420
+ value: 62.104000000000006
421
+ - task:
422
+ type: Retrieval
423
+ dataset:
424
+ type: BeIR/cqadupstack
425
+ name: MTEB CQADupstackGisRetrieval
426
+ config: default
427
+ split: test
428
+ revision: None
429
+ metrics:
430
+ - type: map_at_1
431
+ value: 25.682
432
+ - type: map_at_10
433
+ value: 33.498
434
+ - type: map_at_100
435
+ value: 34.461000000000006
436
+ - type: map_at_1000
437
+ value: 34.544000000000004
438
+ - type: map_at_3
439
+ value: 30.503999999999998
440
+ - type: map_at_5
441
+ value: 32.216
442
+ - type: mrr_at_1
443
+ value: 27.683999999999997
444
+ - type: mrr_at_10
445
+ value: 35.467999999999996
446
+ - type: mrr_at_100
447
+ value: 36.32
448
+ - type: mrr_at_1000
449
+ value: 36.386
450
+ - type: mrr_at_3
451
+ value: 32.618
452
+ - type: mrr_at_5
453
+ value: 34.262
454
+ - type: ndcg_at_1
455
+ value: 27.683999999999997
456
+ - type: ndcg_at_10
457
+ value: 38.378
458
+ - type: ndcg_at_100
459
+ value: 43.288
460
+ - type: ndcg_at_1000
461
+ value: 45.413
462
+ - type: ndcg_at_3
463
+ value: 32.586
464
+ - type: ndcg_at_5
465
+ value: 35.499
466
+ - type: precision_at_1
467
+ value: 27.683999999999997
468
+ - type: precision_at_10
469
+ value: 5.864
470
+ - type: precision_at_100
471
+ value: 0.882
472
+ - type: precision_at_1000
473
+ value: 0.11
474
+ - type: precision_at_3
475
+ value: 13.446
476
+ - type: precision_at_5
477
+ value: 9.718
478
+ - type: recall_at_1
479
+ value: 25.682
480
+ - type: recall_at_10
481
+ value: 51.712
482
+ - type: recall_at_100
483
+ value: 74.446
484
+ - type: recall_at_1000
485
+ value: 90.472
486
+ - type: recall_at_3
487
+ value: 36.236000000000004
488
+ - type: recall_at_5
489
+ value: 43.234
490
+ - task:
491
+ type: Retrieval
492
+ dataset:
493
+ type: BeIR/cqadupstack
494
+ name: MTEB CQADupstackMathematicaRetrieval
495
+ config: default
496
+ split: test
497
+ revision: None
498
+ metrics:
499
+ - type: map_at_1
500
+ value: 16.073999999999998
501
+ - type: map_at_10
502
+ value: 24.352999999999998
503
+ - type: map_at_100
504
+ value: 25.438
505
+ - type: map_at_1000
506
+ value: 25.545
507
+ - type: map_at_3
508
+ value: 21.614
509
+ - type: map_at_5
510
+ value: 23.104
511
+ - type: mrr_at_1
512
+ value: 19.776
513
+ - type: mrr_at_10
514
+ value: 28.837000000000003
515
+ - type: mrr_at_100
516
+ value: 29.755
517
+ - type: mrr_at_1000
518
+ value: 29.817
519
+ - type: mrr_at_3
520
+ value: 26.201999999999998
521
+ - type: mrr_at_5
522
+ value: 27.714
523
+ - type: ndcg_at_1
524
+ value: 19.776
525
+ - type: ndcg_at_10
526
+ value: 29.701
527
+ - type: ndcg_at_100
528
+ value: 35.307
529
+ - type: ndcg_at_1000
530
+ value: 37.942
531
+ - type: ndcg_at_3
532
+ value: 24.764
533
+ - type: ndcg_at_5
534
+ value: 27.025
535
+ - type: precision_at_1
536
+ value: 19.776
537
+ - type: precision_at_10
538
+ value: 5.659
539
+ - type: precision_at_100
540
+ value: 0.971
541
+ - type: precision_at_1000
542
+ value: 0.133
543
+ - type: precision_at_3
544
+ value: 12.065
545
+ - type: precision_at_5
546
+ value: 8.905000000000001
547
+ - type: recall_at_1
548
+ value: 16.073999999999998
549
+ - type: recall_at_10
550
+ value: 41.647
551
+ - type: recall_at_100
552
+ value: 66.884
553
+ - type: recall_at_1000
554
+ value: 85.91499999999999
555
+ - type: recall_at_3
556
+ value: 27.916
557
+ - type: recall_at_5
558
+ value: 33.729
559
+ - task:
560
+ type: Retrieval
561
+ dataset:
562
+ type: BeIR/cqadupstack
563
+ name: MTEB CQADupstackPhysicsRetrieval
564
+ config: default
565
+ split: test
566
+ revision: None
567
+ metrics:
568
+ - type: map_at_1
569
+ value: 28.444999999999997
570
+ - type: map_at_10
571
+ value: 38.218999999999994
572
+ - type: map_at_100
573
+ value: 39.595
574
+ - type: map_at_1000
575
+ value: 39.709
576
+ - type: map_at_3
577
+ value: 35.586
578
+ - type: map_at_5
579
+ value: 36.895
580
+ - type: mrr_at_1
581
+ value: 34.841
582
+ - type: mrr_at_10
583
+ value: 44.106
584
+ - type: mrr_at_100
585
+ value: 44.98
586
+ - type: mrr_at_1000
587
+ value: 45.03
588
+ - type: mrr_at_3
589
+ value: 41.979
590
+ - type: mrr_at_5
591
+ value: 43.047999999999995
592
+ - type: ndcg_at_1
593
+ value: 34.841
594
+ - type: ndcg_at_10
595
+ value: 43.922
596
+ - type: ndcg_at_100
597
+ value: 49.504999999999995
598
+ - type: ndcg_at_1000
599
+ value: 51.675000000000004
600
+ - type: ndcg_at_3
601
+ value: 39.858
602
+ - type: ndcg_at_5
603
+ value: 41.408
604
+ - type: precision_at_1
605
+ value: 34.841
606
+ - type: precision_at_10
607
+ value: 7.872999999999999
608
+ - type: precision_at_100
609
+ value: 1.2449999999999999
610
+ - type: precision_at_1000
611
+ value: 0.161
612
+ - type: precision_at_3
613
+ value: 18.993
614
+ - type: precision_at_5
615
+ value: 13.032
616
+ - type: recall_at_1
617
+ value: 28.444999999999997
618
+ - type: recall_at_10
619
+ value: 54.984
620
+ - type: recall_at_100
621
+ value: 78.342
622
+ - type: recall_at_1000
623
+ value: 92.77
624
+ - type: recall_at_3
625
+ value: 42.842999999999996
626
+ - type: recall_at_5
627
+ value: 47.247
628
+ - task:
629
+ type: Retrieval
630
+ dataset:
631
+ type: BeIR/cqadupstack
632
+ name: MTEB CQADupstackProgrammersRetrieval
633
+ config: default
634
+ split: test
635
+ revision: None
636
+ metrics:
637
+ - type: map_at_1
638
+ value: 23.072
639
+ - type: map_at_10
640
+ value: 32.354
641
+ - type: map_at_100
642
+ value: 33.800000000000004
643
+ - type: map_at_1000
644
+ value: 33.908
645
+ - type: map_at_3
646
+ value: 29.232000000000003
647
+ - type: map_at_5
648
+ value: 31.049
649
+ - type: mrr_at_1
650
+ value: 29.110000000000003
651
+ - type: mrr_at_10
652
+ value: 38.03
653
+ - type: mrr_at_100
654
+ value: 39.032
655
+ - type: mrr_at_1000
656
+ value: 39.086999999999996
657
+ - type: mrr_at_3
658
+ value: 35.407
659
+ - type: mrr_at_5
660
+ value: 36.76
661
+ - type: ndcg_at_1
662
+ value: 29.110000000000003
663
+ - type: ndcg_at_10
664
+ value: 38.231
665
+ - type: ndcg_at_100
666
+ value: 44.425
667
+ - type: ndcg_at_1000
668
+ value: 46.771
669
+ - type: ndcg_at_3
670
+ value: 33.095
671
+ - type: ndcg_at_5
672
+ value: 35.459
673
+ - type: precision_at_1
674
+ value: 29.110000000000003
675
+ - type: precision_at_10
676
+ value: 7.215000000000001
677
+ - type: precision_at_100
678
+ value: 1.2109999999999999
679
+ - type: precision_at_1000
680
+ value: 0.157
681
+ - type: precision_at_3
682
+ value: 16.058
683
+ - type: precision_at_5
684
+ value: 11.644
685
+ - type: recall_at_1
686
+ value: 23.072
687
+ - type: recall_at_10
688
+ value: 50.285999999999994
689
+ - type: recall_at_100
690
+ value: 76.596
691
+ - type: recall_at_1000
692
+ value: 92.861
693
+ - type: recall_at_3
694
+ value: 35.702
695
+ - type: recall_at_5
696
+ value: 42.152
697
+ - task:
698
+ type: Retrieval
699
+ dataset:
700
+ type: BeIR/cqadupstack
701
+ name: MTEB CQADupstackRetrieval
702
+ config: default
703
+ split: test
704
+ revision: None
705
+ metrics:
706
+ - type: map_at_1
707
+ value: 24.937916666666666
708
+ - type: map_at_10
709
+ value: 33.755250000000004
710
+ - type: map_at_100
711
+ value: 34.955999999999996
712
+ - type: map_at_1000
713
+ value: 35.070499999999996
714
+ - type: map_at_3
715
+ value: 30.98708333333333
716
+ - type: map_at_5
717
+ value: 32.51491666666666
718
+ - type: mrr_at_1
719
+ value: 29.48708333333333
720
+ - type: mrr_at_10
721
+ value: 37.92183333333334
722
+ - type: mrr_at_100
723
+ value: 38.76583333333333
724
+ - type: mrr_at_1000
725
+ value: 38.82466666666667
726
+ - type: mrr_at_3
727
+ value: 35.45125
728
+ - type: mrr_at_5
729
+ value: 36.827000000000005
730
+ - type: ndcg_at_1
731
+ value: 29.48708333333333
732
+ - type: ndcg_at_10
733
+ value: 39.05225
734
+ - type: ndcg_at_100
735
+ value: 44.25983333333334
736
+ - type: ndcg_at_1000
737
+ value: 46.568333333333335
738
+ - type: ndcg_at_3
739
+ value: 34.271583333333325
740
+ - type: ndcg_at_5
741
+ value: 36.483916666666666
742
+ - type: precision_at_1
743
+ value: 29.48708333333333
744
+ - type: precision_at_10
745
+ value: 6.865749999999999
746
+ - type: precision_at_100
747
+ value: 1.1195833333333332
748
+ - type: precision_at_1000
749
+ value: 0.15058333333333335
750
+ - type: precision_at_3
751
+ value: 15.742083333333333
752
+ - type: precision_at_5
753
+ value: 11.221916666666667
754
+ - type: recall_at_1
755
+ value: 24.937916666666666
756
+ - type: recall_at_10
757
+ value: 50.650416666666665
758
+ - type: recall_at_100
759
+ value: 73.55383333333334
760
+ - type: recall_at_1000
761
+ value: 89.61691666666667
762
+ - type: recall_at_3
763
+ value: 37.27808333333334
764
+ - type: recall_at_5
765
+ value: 42.99475
766
+ - task:
767
+ type: Retrieval
768
+ dataset:
769
+ type: BeIR/cqadupstack
770
+ name: MTEB CQADupstackStatsRetrieval
771
+ config: default
772
+ split: test
773
+ revision: None
774
+ metrics:
775
+ - type: map_at_1
776
+ value: 23.947
777
+ - type: map_at_10
778
+ value: 30.575000000000003
779
+ - type: map_at_100
780
+ value: 31.465
781
+ - type: map_at_1000
782
+ value: 31.558000000000003
783
+ - type: map_at_3
784
+ value: 28.814
785
+ - type: map_at_5
786
+ value: 29.738999999999997
787
+ - type: mrr_at_1
788
+ value: 26.994
789
+ - type: mrr_at_10
790
+ value: 33.415
791
+ - type: mrr_at_100
792
+ value: 34.18
793
+ - type: mrr_at_1000
794
+ value: 34.245
795
+ - type: mrr_at_3
796
+ value: 31.621
797
+ - type: mrr_at_5
798
+ value: 32.549
799
+ - type: ndcg_at_1
800
+ value: 26.994
801
+ - type: ndcg_at_10
802
+ value: 34.482
803
+ - type: ndcg_at_100
804
+ value: 38.915
805
+ - type: ndcg_at_1000
806
+ value: 41.355
807
+ - type: ndcg_at_3
808
+ value: 31.139
809
+ - type: ndcg_at_5
810
+ value: 32.589
811
+ - type: precision_at_1
812
+ value: 26.994
813
+ - type: precision_at_10
814
+ value: 5.322
815
+ - type: precision_at_100
816
+ value: 0.8160000000000001
817
+ - type: precision_at_1000
818
+ value: 0.11100000000000002
819
+ - type: precision_at_3
820
+ value: 13.344000000000001
821
+ - type: precision_at_5
822
+ value: 8.988
823
+ - type: recall_at_1
824
+ value: 23.947
825
+ - type: recall_at_10
826
+ value: 43.647999999999996
827
+ - type: recall_at_100
828
+ value: 63.851
829
+ - type: recall_at_1000
830
+ value: 82.0
831
+ - type: recall_at_3
832
+ value: 34.288000000000004
833
+ - type: recall_at_5
834
+ value: 38.117000000000004
835
+ - task:
836
+ type: Retrieval
837
+ dataset:
838
+ type: BeIR/cqadupstack
839
+ name: MTEB CQADupstackTexRetrieval
840
+ config: default
841
+ split: test
842
+ revision: None
843
+ metrics:
844
+ - type: map_at_1
845
+ value: 16.197
846
+ - type: map_at_10
847
+ value: 22.968
848
+ - type: map_at_100
849
+ value: 24.095
850
+ - type: map_at_1000
851
+ value: 24.217
852
+ - type: map_at_3
853
+ value: 20.771
854
+ - type: map_at_5
855
+ value: 21.995
856
+ - type: mrr_at_1
857
+ value: 19.511
858
+ - type: mrr_at_10
859
+ value: 26.55
860
+ - type: mrr_at_100
861
+ value: 27.500999999999998
862
+ - type: mrr_at_1000
863
+ value: 27.578999999999997
864
+ - type: mrr_at_3
865
+ value: 24.421
866
+ - type: mrr_at_5
867
+ value: 25.604
868
+ - type: ndcg_at_1
869
+ value: 19.511
870
+ - type: ndcg_at_10
871
+ value: 27.386
872
+ - type: ndcg_at_100
873
+ value: 32.828
874
+ - type: ndcg_at_1000
875
+ value: 35.739
876
+ - type: ndcg_at_3
877
+ value: 23.405
878
+ - type: ndcg_at_5
879
+ value: 25.255
880
+ - type: precision_at_1
881
+ value: 19.511
882
+ - type: precision_at_10
883
+ value: 5.017
884
+ - type: precision_at_100
885
+ value: 0.91
886
+ - type: precision_at_1000
887
+ value: 0.133
888
+ - type: precision_at_3
889
+ value: 11.023
890
+ - type: precision_at_5
891
+ value: 8.025
892
+ - type: recall_at_1
893
+ value: 16.197
894
+ - type: recall_at_10
895
+ value: 37.09
896
+ - type: recall_at_100
897
+ value: 61.778
898
+ - type: recall_at_1000
899
+ value: 82.56599999999999
900
+ - type: recall_at_3
901
+ value: 26.034000000000002
902
+ - type: recall_at_5
903
+ value: 30.762
904
+ - task:
905
+ type: Retrieval
906
+ dataset:
907
+ type: BeIR/cqadupstack
908
+ name: MTEB CQADupstackUnixRetrieval
909
+ config: default
910
+ split: test
911
+ revision: None
912
+ metrics:
913
+ - type: map_at_1
914
+ value: 25.41
915
+ - type: map_at_10
916
+ value: 33.655
917
+ - type: map_at_100
918
+ value: 34.892
919
+ - type: map_at_1000
920
+ value: 34.995
921
+ - type: map_at_3
922
+ value: 30.94
923
+ - type: map_at_5
924
+ value: 32.303
925
+ - type: mrr_at_1
926
+ value: 29.477999999999998
927
+ - type: mrr_at_10
928
+ value: 37.443
929
+ - type: mrr_at_100
930
+ value: 38.383
931
+ - type: mrr_at_1000
932
+ value: 38.440000000000005
933
+ - type: mrr_at_3
934
+ value: 34.949999999999996
935
+ - type: mrr_at_5
936
+ value: 36.228
937
+ - type: ndcg_at_1
938
+ value: 29.477999999999998
939
+ - type: ndcg_at_10
940
+ value: 38.769
941
+ - type: ndcg_at_100
942
+ value: 44.245000000000005
943
+ - type: ndcg_at_1000
944
+ value: 46.593
945
+ - type: ndcg_at_3
946
+ value: 33.623
947
+ - type: ndcg_at_5
948
+ value: 35.766
949
+ - type: precision_at_1
950
+ value: 29.477999999999998
951
+ - type: precision_at_10
952
+ value: 6.455
953
+ - type: precision_at_100
954
+ value: 1.032
955
+ - type: precision_at_1000
956
+ value: 0.135
957
+ - type: precision_at_3
958
+ value: 14.893999999999998
959
+ - type: precision_at_5
960
+ value: 10.485
961
+ - type: recall_at_1
962
+ value: 25.41
963
+ - type: recall_at_10
964
+ value: 50.669
965
+ - type: recall_at_100
966
+ value: 74.084
967
+ - type: recall_at_1000
968
+ value: 90.435
969
+ - type: recall_at_3
970
+ value: 36.679
971
+ - type: recall_at_5
972
+ value: 41.94
973
+ - task:
974
+ type: Retrieval
975
+ dataset:
976
+ type: BeIR/cqadupstack
977
+ name: MTEB CQADupstackWebmastersRetrieval
978
+ config: default
979
+ split: test
980
+ revision: None
981
+ metrics:
982
+ - type: map_at_1
983
+ value: 23.339
984
+ - type: map_at_10
985
+ value: 31.852000000000004
986
+ - type: map_at_100
987
+ value: 33.411
988
+ - type: map_at_1000
989
+ value: 33.62
990
+ - type: map_at_3
991
+ value: 28.929
992
+ - type: map_at_5
993
+ value: 30.542
994
+ - type: mrr_at_1
995
+ value: 28.063
996
+ - type: mrr_at_10
997
+ value: 36.301
998
+ - type: mrr_at_100
999
+ value: 37.288
1000
+ - type: mrr_at_1000
1001
+ value: 37.349
1002
+ - type: mrr_at_3
1003
+ value: 33.663
1004
+ - type: mrr_at_5
1005
+ value: 35.165
1006
+ - type: ndcg_at_1
1007
+ value: 28.063
1008
+ - type: ndcg_at_10
1009
+ value: 37.462
1010
+ - type: ndcg_at_100
1011
+ value: 43.620999999999995
1012
+ - type: ndcg_at_1000
1013
+ value: 46.211
1014
+ - type: ndcg_at_3
1015
+ value: 32.68
1016
+ - type: ndcg_at_5
1017
+ value: 34.981
1018
+ - type: precision_at_1
1019
+ value: 28.063
1020
+ - type: precision_at_10
1021
+ value: 7.1739999999999995
1022
+ - type: precision_at_100
1023
+ value: 1.486
1024
+ - type: precision_at_1000
1025
+ value: 0.23500000000000001
1026
+ - type: precision_at_3
1027
+ value: 15.217
1028
+ - type: precision_at_5
1029
+ value: 11.265
1030
+ - type: recall_at_1
1031
+ value: 23.339
1032
+ - type: recall_at_10
1033
+ value: 48.376999999999995
1034
+ - type: recall_at_100
1035
+ value: 76.053
1036
+ - type: recall_at_1000
1037
+ value: 92.455
1038
+ - type: recall_at_3
1039
+ value: 34.735
1040
+ - type: recall_at_5
1041
+ value: 40.71
1042
+ - task:
1043
+ type: Retrieval
1044
+ dataset:
1045
+ type: BeIR/cqadupstack
1046
+ name: MTEB CQADupstackWordpressRetrieval
1047
+ config: default
1048
+ split: test
1049
+ revision: None
1050
+ metrics:
1051
+ - type: map_at_1
1052
+ value: 18.925
1053
+ - type: map_at_10
1054
+ value: 26.017000000000003
1055
+ - type: map_at_100
1056
+ value: 27.034000000000002
1057
+ - type: map_at_1000
1058
+ value: 27.156000000000002
1059
+ - type: map_at_3
1060
+ value: 23.604
1061
+ - type: map_at_5
1062
+ value: 24.75
1063
+ - type: mrr_at_1
1064
+ value: 20.333000000000002
1065
+ - type: mrr_at_10
1066
+ value: 27.915
1067
+ - type: mrr_at_100
1068
+ value: 28.788000000000004
1069
+ - type: mrr_at_1000
1070
+ value: 28.877999999999997
1071
+ - type: mrr_at_3
1072
+ value: 25.446999999999996
1073
+ - type: mrr_at_5
1074
+ value: 26.648
1075
+ - type: ndcg_at_1
1076
+ value: 20.333000000000002
1077
+ - type: ndcg_at_10
1078
+ value: 30.673000000000002
1079
+ - type: ndcg_at_100
1080
+ value: 35.618
1081
+ - type: ndcg_at_1000
1082
+ value: 38.517
1083
+ - type: ndcg_at_3
1084
+ value: 25.71
1085
+ - type: ndcg_at_5
1086
+ value: 27.679
1087
+ - type: precision_at_1
1088
+ value: 20.333000000000002
1089
+ - type: precision_at_10
1090
+ value: 4.9910000000000005
1091
+ - type: precision_at_100
1092
+ value: 0.8130000000000001
1093
+ - type: precision_at_1000
1094
+ value: 0.117
1095
+ - type: precision_at_3
1096
+ value: 11.029
1097
+ - type: precision_at_5
1098
+ value: 7.8740000000000006
1099
+ - type: recall_at_1
1100
+ value: 18.925
1101
+ - type: recall_at_10
1102
+ value: 43.311
1103
+ - type: recall_at_100
1104
+ value: 66.308
1105
+ - type: recall_at_1000
1106
+ value: 87.49
1107
+ - type: recall_at_3
1108
+ value: 29.596
1109
+ - type: recall_at_5
1110
+ value: 34.245
1111
+ - task:
1112
+ type: Retrieval
1113
+ dataset:
1114
+ type: climate-fever
1115
+ name: MTEB ClimateFEVER
1116
+ config: default
1117
+ split: test
1118
+ revision: None
1119
+ metrics:
1120
+ - type: map_at_1
1121
+ value: 13.714
1122
+ - type: map_at_10
1123
+ value: 23.194
1124
+ - type: map_at_100
1125
+ value: 24.976000000000003
1126
+ - type: map_at_1000
1127
+ value: 25.166
1128
+ - type: map_at_3
1129
+ value: 19.709
1130
+ - type: map_at_5
1131
+ value: 21.523999999999997
1132
+ - type: mrr_at_1
1133
+ value: 30.619000000000003
1134
+ - type: mrr_at_10
1135
+ value: 42.563
1136
+ - type: mrr_at_100
1137
+ value: 43.386
1138
+ - type: mrr_at_1000
1139
+ value: 43.423
1140
+ - type: mrr_at_3
1141
+ value: 39.555
1142
+ - type: mrr_at_5
1143
+ value: 41.268
1144
+ - type: ndcg_at_1
1145
+ value: 30.619000000000003
1146
+ - type: ndcg_at_10
1147
+ value: 31.836
1148
+ - type: ndcg_at_100
1149
+ value: 38.652
1150
+ - type: ndcg_at_1000
1151
+ value: 42.088
1152
+ - type: ndcg_at_3
1153
+ value: 26.733
1154
+ - type: ndcg_at_5
1155
+ value: 28.435
1156
+ - type: precision_at_1
1157
+ value: 30.619000000000003
1158
+ - type: precision_at_10
1159
+ value: 9.751999999999999
1160
+ - type: precision_at_100
1161
+ value: 1.71
1162
+ - type: precision_at_1000
1163
+ value: 0.23500000000000001
1164
+ - type: precision_at_3
1165
+ value: 19.935
1166
+ - type: precision_at_5
1167
+ value: 14.984
1168
+ - type: recall_at_1
1169
+ value: 13.714
1170
+ - type: recall_at_10
1171
+ value: 37.26
1172
+ - type: recall_at_100
1173
+ value: 60.546
1174
+ - type: recall_at_1000
1175
+ value: 79.899
1176
+ - type: recall_at_3
1177
+ value: 24.325
1178
+ - type: recall_at_5
1179
+ value: 29.725
1180
+ - task:
1181
+ type: Retrieval
1182
+ dataset:
1183
+ type: dbpedia-entity
1184
+ name: MTEB DBPedia
1185
+ config: default
1186
+ split: test
1187
+ revision: None
1188
+ metrics:
1189
+ - type: map_at_1
1190
+ value: 8.462
1191
+ - type: map_at_10
1192
+ value: 18.637
1193
+ - type: map_at_100
1194
+ value: 26.131999999999998
1195
+ - type: map_at_1000
1196
+ value: 27.607
1197
+ - type: map_at_3
1198
+ value: 13.333
1199
+ - type: map_at_5
1200
+ value: 15.654000000000002
1201
+ - type: mrr_at_1
1202
+ value: 66.25
1203
+ - type: mrr_at_10
1204
+ value: 74.32600000000001
1205
+ - type: mrr_at_100
1206
+ value: 74.60900000000001
1207
+ - type: mrr_at_1000
1208
+ value: 74.62
1209
+ - type: mrr_at_3
1210
+ value: 72.667
1211
+ - type: mrr_at_5
1212
+ value: 73.817
1213
+ - type: ndcg_at_1
1214
+ value: 53.87499999999999
1215
+ - type: ndcg_at_10
1216
+ value: 40.028999999999996
1217
+ - type: ndcg_at_100
1218
+ value: 44.199
1219
+ - type: ndcg_at_1000
1220
+ value: 51.629999999999995
1221
+ - type: ndcg_at_3
1222
+ value: 44.113
1223
+ - type: ndcg_at_5
1224
+ value: 41.731
1225
+ - type: precision_at_1
1226
+ value: 66.25
1227
+ - type: precision_at_10
1228
+ value: 31.900000000000002
1229
+ - type: precision_at_100
1230
+ value: 10.043000000000001
1231
+ - type: precision_at_1000
1232
+ value: 1.926
1233
+ - type: precision_at_3
1234
+ value: 47.417
1235
+ - type: precision_at_5
1236
+ value: 40.65
1237
+ - type: recall_at_1
1238
+ value: 8.462
1239
+ - type: recall_at_10
1240
+ value: 24.293
1241
+ - type: recall_at_100
1242
+ value: 50.146
1243
+ - type: recall_at_1000
1244
+ value: 74.034
1245
+ - type: recall_at_3
1246
+ value: 14.967
1247
+ - type: recall_at_5
1248
+ value: 18.682000000000002
1249
+ - task:
1250
+ type: Classification
1251
+ dataset:
1252
+ type: mteb/emotion
1253
+ name: MTEB EmotionClassification
1254
+ config: default
1255
+ split: test
1256
+ revision: 4f58c6b202a23cf9a4da393831edf4f9183cad37
1257
+ metrics:
1258
+ - type: accuracy
1259
+ value: 47.84499999999999
1260
+ - type: f1
1261
+ value: 42.48106691979349
1262
+ - task:
1263
+ type: Retrieval
1264
+ dataset:
1265
+ type: fever
1266
+ name: MTEB FEVER
1267
+ config: default
1268
+ split: test
1269
+ revision: None
1270
+ metrics:
1271
+ - type: map_at_1
1272
+ value: 74.034
1273
+ - type: map_at_10
1274
+ value: 82.76
1275
+ - type: map_at_100
1276
+ value: 82.968
1277
+ - type: map_at_1000
1278
+ value: 82.98299999999999
1279
+ - type: map_at_3
1280
+ value: 81.768
1281
+ - type: map_at_5
1282
+ value: 82.418
1283
+ - type: mrr_at_1
1284
+ value: 80.048
1285
+ - type: mrr_at_10
1286
+ value: 87.64999999999999
1287
+ - type: mrr_at_100
1288
+ value: 87.712
1289
+ - type: mrr_at_1000
1290
+ value: 87.713
1291
+ - type: mrr_at_3
1292
+ value: 87.01100000000001
1293
+ - type: mrr_at_5
1294
+ value: 87.466
1295
+ - type: ndcg_at_1
1296
+ value: 80.048
1297
+ - type: ndcg_at_10
1298
+ value: 86.643
1299
+ - type: ndcg_at_100
1300
+ value: 87.361
1301
+ - type: ndcg_at_1000
1302
+ value: 87.606
1303
+ - type: ndcg_at_3
1304
+ value: 85.137
1305
+ - type: ndcg_at_5
1306
+ value: 86.016
1307
+ - type: precision_at_1
1308
+ value: 80.048
1309
+ - type: precision_at_10
1310
+ value: 10.372
1311
+ - type: precision_at_100
1312
+ value: 1.093
1313
+ - type: precision_at_1000
1314
+ value: 0.11299999999999999
1315
+ - type: precision_at_3
1316
+ value: 32.638
1317
+ - type: precision_at_5
1318
+ value: 20.177
1319
+ - type: recall_at_1
1320
+ value: 74.034
1321
+ - type: recall_at_10
1322
+ value: 93.769
1323
+ - type: recall_at_100
1324
+ value: 96.569
1325
+ - type: recall_at_1000
1326
+ value: 98.039
1327
+ - type: recall_at_3
1328
+ value: 89.581
1329
+ - type: recall_at_5
1330
+ value: 91.906
1331
+ - task:
1332
+ type: Retrieval
1333
+ dataset:
1334
+ type: fiqa
1335
+ name: MTEB FiQA2018
1336
+ config: default
1337
+ split: test
1338
+ revision: None
1339
+ metrics:
1340
+ - type: map_at_1
1341
+ value: 20.5
1342
+ - type: map_at_10
1343
+ value: 32.857
1344
+ - type: map_at_100
1345
+ value: 34.589
1346
+ - type: map_at_1000
1347
+ value: 34.778
1348
+ - type: map_at_3
1349
+ value: 29.160999999999998
1350
+ - type: map_at_5
1351
+ value: 31.033
1352
+ - type: mrr_at_1
1353
+ value: 40.123
1354
+ - type: mrr_at_10
1355
+ value: 48.776
1356
+ - type: mrr_at_100
1357
+ value: 49.495
1358
+ - type: mrr_at_1000
1359
+ value: 49.539
1360
+ - type: mrr_at_3
1361
+ value: 46.605000000000004
1362
+ - type: mrr_at_5
1363
+ value: 47.654
1364
+ - type: ndcg_at_1
1365
+ value: 40.123
1366
+ - type: ndcg_at_10
1367
+ value: 40.343
1368
+ - type: ndcg_at_100
1369
+ value: 46.56
1370
+ - type: ndcg_at_1000
1371
+ value: 49.777
1372
+ - type: ndcg_at_3
1373
+ value: 37.322
1374
+ - type: ndcg_at_5
1375
+ value: 37.791000000000004
1376
+ - type: precision_at_1
1377
+ value: 40.123
1378
+ - type: precision_at_10
1379
+ value: 11.08
1380
+ - type: precision_at_100
1381
+ value: 1.752
1382
+ - type: precision_at_1000
1383
+ value: 0.232
1384
+ - type: precision_at_3
1385
+ value: 24.897
1386
+ - type: precision_at_5
1387
+ value: 17.809
1388
+ - type: recall_at_1
1389
+ value: 20.5
1390
+ - type: recall_at_10
1391
+ value: 46.388
1392
+ - type: recall_at_100
1393
+ value: 69.552
1394
+ - type: recall_at_1000
1395
+ value: 89.011
1396
+ - type: recall_at_3
1397
+ value: 33.617999999999995
1398
+ - type: recall_at_5
1399
+ value: 38.211
1400
+ - task:
1401
+ type: Retrieval
1402
+ dataset:
1403
+ type: hotpotqa
1404
+ name: MTEB HotpotQA
1405
+ config: default
1406
+ split: test
1407
+ revision: None
1408
+ metrics:
1409
+ - type: map_at_1
1410
+ value: 39.135999999999996
1411
+ - type: map_at_10
1412
+ value: 61.673
1413
+ - type: map_at_100
1414
+ value: 62.562
1415
+ - type: map_at_1000
1416
+ value: 62.62
1417
+ - type: map_at_3
1418
+ value: 58.467999999999996
1419
+ - type: map_at_5
1420
+ value: 60.463
1421
+ - type: mrr_at_1
1422
+ value: 78.271
1423
+ - type: mrr_at_10
1424
+ value: 84.119
1425
+ - type: mrr_at_100
1426
+ value: 84.29299999999999
1427
+ - type: mrr_at_1000
1428
+ value: 84.299
1429
+ - type: mrr_at_3
1430
+ value: 83.18900000000001
1431
+ - type: mrr_at_5
1432
+ value: 83.786
1433
+ - type: ndcg_at_1
1434
+ value: 78.271
1435
+ - type: ndcg_at_10
1436
+ value: 69.935
1437
+ - type: ndcg_at_100
1438
+ value: 73.01299999999999
1439
+ - type: ndcg_at_1000
1440
+ value: 74.126
1441
+ - type: ndcg_at_3
1442
+ value: 65.388
1443
+ - type: ndcg_at_5
1444
+ value: 67.906
1445
+ - type: precision_at_1
1446
+ value: 78.271
1447
+ - type: precision_at_10
1448
+ value: 14.562
1449
+ - type: precision_at_100
1450
+ value: 1.6969999999999998
1451
+ - type: precision_at_1000
1452
+ value: 0.184
1453
+ - type: precision_at_3
1454
+ value: 41.841
1455
+ - type: precision_at_5
1456
+ value: 27.087
1457
+ - type: recall_at_1
1458
+ value: 39.135999999999996
1459
+ - type: recall_at_10
1460
+ value: 72.809
1461
+ - type: recall_at_100
1462
+ value: 84.86200000000001
1463
+ - type: recall_at_1000
1464
+ value: 92.208
1465
+ - type: recall_at_3
1466
+ value: 62.76199999999999
1467
+ - type: recall_at_5
1468
+ value: 67.718
1469
+ - task:
1470
+ type: Classification
1471
+ dataset:
1472
+ type: mteb/imdb
1473
+ name: MTEB ImdbClassification
1474
+ config: default
1475
+ split: test
1476
+ revision: 3d86128a09e091d6018b6d26cad27f2739fc2db7
1477
+ metrics:
1478
+ - type: accuracy
1479
+ value: 90.60600000000001
1480
+ - type: ap
1481
+ value: 86.6579587804335
1482
+ - type: f1
1483
+ value: 90.5938853929307
1484
+ - task:
1485
+ type: Retrieval
1486
+ dataset:
1487
+ type: msmarco
1488
+ name: MTEB MSMARCO
1489
+ config: default
1490
+ split: dev
1491
+ revision: None
1492
+ metrics:
1493
+ - type: map_at_1
1494
+ value: 21.852
1495
+ - type: map_at_10
1496
+ value: 33.982
1497
+ - type: map_at_100
1498
+ value: 35.116
1499
+ - type: map_at_1000
1500
+ value: 35.167
1501
+ - type: map_at_3
1502
+ value: 30.134
1503
+ - type: map_at_5
1504
+ value: 32.340999999999994
1505
+ - type: mrr_at_1
1506
+ value: 22.479
1507
+ - type: mrr_at_10
1508
+ value: 34.594
1509
+ - type: mrr_at_100
1510
+ value: 35.672
1511
+ - type: mrr_at_1000
1512
+ value: 35.716
1513
+ - type: mrr_at_3
1514
+ value: 30.84
1515
+ - type: mrr_at_5
1516
+ value: 32.998
1517
+ - type: ndcg_at_1
1518
+ value: 22.493
1519
+ - type: ndcg_at_10
1520
+ value: 40.833000000000006
1521
+ - type: ndcg_at_100
1522
+ value: 46.357
1523
+ - type: ndcg_at_1000
1524
+ value: 47.637
1525
+ - type: ndcg_at_3
1526
+ value: 32.995999999999995
1527
+ - type: ndcg_at_5
1528
+ value: 36.919000000000004
1529
+ - type: precision_at_1
1530
+ value: 22.493
1531
+ - type: precision_at_10
1532
+ value: 6.465999999999999
1533
+ - type: precision_at_100
1534
+ value: 0.9249999999999999
1535
+ - type: precision_at_1000
1536
+ value: 0.104
1537
+ - type: precision_at_3
1538
+ value: 14.030999999999999
1539
+ - type: precision_at_5
1540
+ value: 10.413
1541
+ - type: recall_at_1
1542
+ value: 21.852
1543
+ - type: recall_at_10
1544
+ value: 61.934999999999995
1545
+ - type: recall_at_100
1546
+ value: 87.611
1547
+ - type: recall_at_1000
1548
+ value: 97.441
1549
+ - type: recall_at_3
1550
+ value: 40.583999999999996
1551
+ - type: recall_at_5
1552
+ value: 49.992999999999995
1553
+ - task:
1554
+ type: Classification
1555
+ dataset:
1556
+ type: mteb/mtop_domain
1557
+ name: MTEB MTOPDomainClassification (en)
1558
+ config: en
1559
+ split: test
1560
+ revision: d80d48c1eb48d3562165c59d59d0034df9fff0bf
1561
+ metrics:
1562
+ - type: accuracy
1563
+ value: 93.36069311445507
1564
+ - type: f1
1565
+ value: 93.16456330371453
1566
+ - task:
1567
+ type: Classification
1568
+ dataset:
1569
+ type: mteb/mtop_intent
1570
+ name: MTEB MTOPIntentClassification (en)
1571
+ config: en
1572
+ split: test
1573
+ revision: ae001d0e6b1228650b7bd1c2c65fb50ad11a8aba
1574
+ metrics:
1575
+ - type: accuracy
1576
+ value: 74.74692202462381
1577
+ - type: f1
1578
+ value: 58.17903579421599
1579
+ - task:
1580
+ type: Classification
1581
+ dataset:
1582
+ type: mteb/amazon_massive_intent
1583
+ name: MTEB MassiveIntentClassification (en)
1584
+ config: en
1585
+ split: test
1586
+ revision: 31efe3c427b0bae9c22cbb560b8f15491cc6bed7
1587
+ metrics:
1588
+ - type: accuracy
1589
+ value: 74.80833893745796
1590
+ - type: f1
1591
+ value: 72.70786592684664
1592
+ - task:
1593
+ type: Classification
1594
+ dataset:
1595
+ type: mteb/amazon_massive_scenario
1596
+ name: MTEB MassiveScenarioClassification (en)
1597
+ config: en
1598
+ split: test
1599
+ revision: 7d571f92784cd94a019292a1f45445077d0ef634
1600
+ metrics:
1601
+ - type: accuracy
1602
+ value: 78.69872225958305
1603
+ - type: f1
1604
+ value: 78.61626934504731
1605
+ - task:
1606
+ type: Clustering
1607
+ dataset:
1608
+ type: mteb/medrxiv-clustering-p2p
1609
+ name: MTEB MedrxivClusteringP2P
1610
+ config: default
1611
+ split: test
1612
+ revision: e7a26af6f3ae46b30dde8737f02c07b1505bcc73
1613
+ metrics:
1614
+ - type: v_measure
1615
+ value: 33.058658628717694
1616
+ - task:
1617
+ type: Clustering
1618
+ dataset:
1619
+ type: mteb/medrxiv-clustering-s2s
1620
+ name: MTEB MedrxivClusteringS2S
1621
+ config: default
1622
+ split: test
1623
+ revision: 35191c8c0dca72d8ff3efcd72aa802307d469663
1624
+ metrics:
1625
+ - type: v_measure
1626
+ value: 30.85561739360599
1627
+ - task:
1628
+ type: Reranking
1629
+ dataset:
1630
+ type: mteb/mind_small
1631
+ name: MTEB MindSmallReranking
1632
+ config: default
1633
+ split: test
1634
+ revision: 3bdac13927fdc888b903db93b2ffdbd90b295a69
1635
+ metrics:
1636
+ - type: map
1637
+ value: 31.290259910144385
1638
+ - type: mrr
1639
+ value: 32.44223046102856
1640
+ - task:
1641
+ type: Retrieval
1642
+ dataset:
1643
+ type: nfcorpus
1644
+ name: MTEB NFCorpus
1645
+ config: default
1646
+ split: test
1647
+ revision: None
1648
+ metrics:
1649
+ - type: map_at_1
1650
+ value: 5.288
1651
+ - type: map_at_10
1652
+ value: 12.267999999999999
1653
+ - type: map_at_100
1654
+ value: 15.557000000000002
1655
+ - type: map_at_1000
1656
+ value: 16.98
1657
+ - type: map_at_3
1658
+ value: 8.866
1659
+ - type: map_at_5
1660
+ value: 10.418
1661
+ - type: mrr_at_1
1662
+ value: 43.653
1663
+ - type: mrr_at_10
1664
+ value: 52.681
1665
+ - type: mrr_at_100
1666
+ value: 53.315999999999995
1667
+ - type: mrr_at_1000
1668
+ value: 53.357
1669
+ - type: mrr_at_3
1670
+ value: 51.393
1671
+ - type: mrr_at_5
1672
+ value: 51.903999999999996
1673
+ - type: ndcg_at_1
1674
+ value: 42.415000000000006
1675
+ - type: ndcg_at_10
1676
+ value: 34.305
1677
+ - type: ndcg_at_100
1678
+ value: 30.825999999999997
1679
+ - type: ndcg_at_1000
1680
+ value: 39.393
1681
+ - type: ndcg_at_3
1682
+ value: 39.931
1683
+ - type: ndcg_at_5
1684
+ value: 37.519999999999996
1685
+ - type: precision_at_1
1686
+ value: 43.653
1687
+ - type: precision_at_10
1688
+ value: 25.728
1689
+ - type: precision_at_100
1690
+ value: 7.932
1691
+ - type: precision_at_1000
1692
+ value: 2.07
1693
+ - type: precision_at_3
1694
+ value: 38.184000000000005
1695
+ - type: precision_at_5
1696
+ value: 32.879000000000005
1697
+ - type: recall_at_1
1698
+ value: 5.288
1699
+ - type: recall_at_10
1700
+ value: 16.195
1701
+ - type: recall_at_100
1702
+ value: 31.135
1703
+ - type: recall_at_1000
1704
+ value: 61.531000000000006
1705
+ - type: recall_at_3
1706
+ value: 10.313
1707
+ - type: recall_at_5
1708
+ value: 12.754999999999999
1709
+ - task:
1710
+ type: Retrieval
1711
+ dataset:
1712
+ type: nq
1713
+ name: MTEB NQ
1714
+ config: default
1715
+ split: test
1716
+ revision: None
1717
+ metrics:
1718
+ - type: map_at_1
1719
+ value: 28.216
1720
+ - type: map_at_10
1721
+ value: 42.588
1722
+ - type: map_at_100
1723
+ value: 43.702999999999996
1724
+ - type: map_at_1000
1725
+ value: 43.739
1726
+ - type: map_at_3
1727
+ value: 38.177
1728
+ - type: map_at_5
1729
+ value: 40.754000000000005
1730
+ - type: mrr_at_1
1731
+ value: 31.866
1732
+ - type: mrr_at_10
1733
+ value: 45.189
1734
+ - type: mrr_at_100
1735
+ value: 46.056000000000004
1736
+ - type: mrr_at_1000
1737
+ value: 46.081
1738
+ - type: mrr_at_3
1739
+ value: 41.526999999999994
1740
+ - type: mrr_at_5
1741
+ value: 43.704
1742
+ - type: ndcg_at_1
1743
+ value: 31.837
1744
+ - type: ndcg_at_10
1745
+ value: 50.178
1746
+ - type: ndcg_at_100
1747
+ value: 54.98800000000001
1748
+ - type: ndcg_at_1000
1749
+ value: 55.812
1750
+ - type: ndcg_at_3
1751
+ value: 41.853
1752
+ - type: ndcg_at_5
1753
+ value: 46.153
1754
+ - type: precision_at_1
1755
+ value: 31.837
1756
+ - type: precision_at_10
1757
+ value: 8.43
1758
+ - type: precision_at_100
1759
+ value: 1.1119999999999999
1760
+ - type: precision_at_1000
1761
+ value: 0.11900000000000001
1762
+ - type: precision_at_3
1763
+ value: 19.023
1764
+ - type: precision_at_5
1765
+ value: 13.911000000000001
1766
+ - type: recall_at_1
1767
+ value: 28.216
1768
+ - type: recall_at_10
1769
+ value: 70.8
1770
+ - type: recall_at_100
1771
+ value: 91.857
1772
+ - type: recall_at_1000
1773
+ value: 97.941
1774
+ - type: recall_at_3
1775
+ value: 49.196
1776
+ - type: recall_at_5
1777
+ value: 59.072
1778
+ - task:
1779
+ type: Retrieval
1780
+ dataset:
1781
+ type: quora
1782
+ name: MTEB QuoraRetrieval
1783
+ config: default
1784
+ split: test
1785
+ revision: None
1786
+ metrics:
1787
+ - type: map_at_1
1788
+ value: 71.22800000000001
1789
+ - type: map_at_10
1790
+ value: 85.115
1791
+ - type: map_at_100
1792
+ value: 85.72
1793
+ - type: map_at_1000
1794
+ value: 85.737
1795
+ - type: map_at_3
1796
+ value: 82.149
1797
+ - type: map_at_5
1798
+ value: 84.029
1799
+ - type: mrr_at_1
1800
+ value: 81.96
1801
+ - type: mrr_at_10
1802
+ value: 88.00200000000001
1803
+ - type: mrr_at_100
1804
+ value: 88.088
1805
+ - type: mrr_at_1000
1806
+ value: 88.089
1807
+ - type: mrr_at_3
1808
+ value: 87.055
1809
+ - type: mrr_at_5
1810
+ value: 87.715
1811
+ - type: ndcg_at_1
1812
+ value: 82.01
1813
+ - type: ndcg_at_10
1814
+ value: 88.78
1815
+ - type: ndcg_at_100
1816
+ value: 89.91
1817
+ - type: ndcg_at_1000
1818
+ value: 90.013
1819
+ - type: ndcg_at_3
1820
+ value: 85.957
1821
+ - type: ndcg_at_5
1822
+ value: 87.56
1823
+ - type: precision_at_1
1824
+ value: 82.01
1825
+ - type: precision_at_10
1826
+ value: 13.462
1827
+ - type: precision_at_100
1828
+ value: 1.528
1829
+ - type: precision_at_1000
1830
+ value: 0.157
1831
+ - type: precision_at_3
1832
+ value: 37.553
1833
+ - type: precision_at_5
1834
+ value: 24.732000000000003
1835
+ - type: recall_at_1
1836
+ value: 71.22800000000001
1837
+ - type: recall_at_10
1838
+ value: 95.69
1839
+ - type: recall_at_100
1840
+ value: 99.531
1841
+ - type: recall_at_1000
1842
+ value: 99.98
1843
+ - type: recall_at_3
1844
+ value: 87.632
1845
+ - type: recall_at_5
1846
+ value: 92.117
1847
+ - task:
1848
+ type: Clustering
1849
+ dataset:
1850
+ type: mteb/reddit-clustering
1851
+ name: MTEB RedditClustering
1852
+ config: default
1853
+ split: test
1854
+ revision: 24640382cdbf8abc73003fb0fa6d111a705499eb
1855
+ metrics:
1856
+ - type: v_measure
1857
+ value: 52.31768034366916
1858
+ - task:
1859
+ type: Clustering
1860
+ dataset:
1861
+ type: mteb/reddit-clustering-p2p
1862
+ name: MTEB RedditClusteringP2P
1863
+ config: default
1864
+ split: test
1865
+ revision: 282350215ef01743dc01b456c7f5241fa8937f16
1866
+ metrics:
1867
+ - type: v_measure
1868
+ value: 60.640266772723606
1869
+ - task:
1870
+ type: Retrieval
1871
+ dataset:
1872
+ type: scidocs
1873
+ name: MTEB SCIDOCS
1874
+ config: default
1875
+ split: test
1876
+ revision: None
1877
+ metrics:
1878
+ - type: map_at_1
1879
+ value: 4.7780000000000005
1880
+ - type: map_at_10
1881
+ value: 12.299
1882
+ - type: map_at_100
1883
+ value: 14.363000000000001
1884
+ - type: map_at_1000
1885
+ value: 14.71
1886
+ - type: map_at_3
1887
+ value: 8.738999999999999
1888
+ - type: map_at_5
1889
+ value: 10.397
1890
+ - type: mrr_at_1
1891
+ value: 23.599999999999998
1892
+ - type: mrr_at_10
1893
+ value: 34.845
1894
+ - type: mrr_at_100
1895
+ value: 35.916
1896
+ - type: mrr_at_1000
1897
+ value: 35.973
1898
+ - type: mrr_at_3
1899
+ value: 31.7
1900
+ - type: mrr_at_5
1901
+ value: 33.535
1902
+ - type: ndcg_at_1
1903
+ value: 23.599999999999998
1904
+ - type: ndcg_at_10
1905
+ value: 20.522000000000002
1906
+ - type: ndcg_at_100
1907
+ value: 28.737000000000002
1908
+ - type: ndcg_at_1000
1909
+ value: 34.596
1910
+ - type: ndcg_at_3
1911
+ value: 19.542
1912
+ - type: ndcg_at_5
1913
+ value: 16.958000000000002
1914
+ - type: precision_at_1
1915
+ value: 23.599999999999998
1916
+ - type: precision_at_10
1917
+ value: 10.67
1918
+ - type: precision_at_100
1919
+ value: 2.259
1920
+ - type: precision_at_1000
1921
+ value: 0.367
1922
+ - type: precision_at_3
1923
+ value: 18.333
1924
+ - type: precision_at_5
1925
+ value: 14.879999999999999
1926
+ - type: recall_at_1
1927
+ value: 4.7780000000000005
1928
+ - type: recall_at_10
1929
+ value: 21.617
1930
+ - type: recall_at_100
1931
+ value: 45.905
1932
+ - type: recall_at_1000
1933
+ value: 74.42
1934
+ - type: recall_at_3
1935
+ value: 11.148
1936
+ - type: recall_at_5
1937
+ value: 15.082999999999998
1938
+ - task:
1939
+ type: STS
1940
+ dataset:
1941
+ type: mteb/sickr-sts
1942
+ name: MTEB SICK-R
1943
+ config: default
1944
+ split: test
1945
+ revision: a6ea5a8cab320b040a23452cc28066d9beae2cee
1946
+ metrics:
1947
+ - type: cos_sim_pearson
1948
+ value: 83.22372750297885
1949
+ - type: cos_sim_spearman
1950
+ value: 79.40972617119405
1951
+ - type: euclidean_pearson
1952
+ value: 80.6101072020434
1953
+ - type: euclidean_spearman
1954
+ value: 79.53844217225202
1955
+ - type: manhattan_pearson
1956
+ value: 80.57265975286111
1957
+ - type: manhattan_spearman
1958
+ value: 79.46335611792958
1959
+ - task:
1960
+ type: STS
1961
+ dataset:
1962
+ type: mteb/sts12-sts
1963
+ name: MTEB STS12
1964
+ config: default
1965
+ split: test
1966
+ revision: a0d554a64d88156834ff5ae9920b964011b16384
1967
+ metrics:
1968
+ - type: cos_sim_pearson
1969
+ value: 85.43713315520749
1970
+ - type: cos_sim_spearman
1971
+ value: 77.44128693329532
1972
+ - type: euclidean_pearson
1973
+ value: 81.63869928101123
1974
+ - type: euclidean_spearman
1975
+ value: 77.29512977961515
1976
+ - type: manhattan_pearson
1977
+ value: 81.63704185566183
1978
+ - type: manhattan_spearman
1979
+ value: 77.29909412738657
1980
+ - task:
1981
+ type: STS
1982
+ dataset:
1983
+ type: mteb/sts13-sts
1984
+ name: MTEB STS13
1985
+ config: default
1986
+ split: test
1987
+ revision: 7e90230a92c190f1bf69ae9002b8cea547a64cca
1988
+ metrics:
1989
+ - type: cos_sim_pearson
1990
+ value: 81.59451537860527
1991
+ - type: cos_sim_spearman
1992
+ value: 82.97994638856723
1993
+ - type: euclidean_pearson
1994
+ value: 82.89478688288412
1995
+ - type: euclidean_spearman
1996
+ value: 83.58740751053104
1997
+ - type: manhattan_pearson
1998
+ value: 82.69140840941608
1999
+ - type: manhattan_spearman
2000
+ value: 83.33665956040555
2001
+ - task:
2002
+ type: STS
2003
+ dataset:
2004
+ type: mteb/sts14-sts
2005
+ name: MTEB STS14
2006
+ config: default
2007
+ split: test
2008
+ revision: 6031580fec1f6af667f0bd2da0a551cf4f0b2375
2009
+ metrics:
2010
+ - type: cos_sim_pearson
2011
+ value: 82.00756527711764
2012
+ - type: cos_sim_spearman
2013
+ value: 81.83560996841379
2014
+ - type: euclidean_pearson
2015
+ value: 82.07684151976518
2016
+ - type: euclidean_spearman
2017
+ value: 82.00913052060511
2018
+ - type: manhattan_pearson
2019
+ value: 82.05690778488794
2020
+ - type: manhattan_spearman
2021
+ value: 82.02260252019525
2022
+ - task:
2023
+ type: STS
2024
+ dataset:
2025
+ type: mteb/sts15-sts
2026
+ name: MTEB STS15
2027
+ config: default
2028
+ split: test
2029
+ revision: ae752c7c21bf194d8b67fd573edf7ae58183cbe3
2030
+ metrics:
2031
+ - type: cos_sim_pearson
2032
+ value: 86.13710262895447
2033
+ - type: cos_sim_spearman
2034
+ value: 87.26412811156248
2035
+ - type: euclidean_pearson
2036
+ value: 86.94151453230228
2037
+ - type: euclidean_spearman
2038
+ value: 87.5363796699571
2039
+ - type: manhattan_pearson
2040
+ value: 86.86989424083748
2041
+ - type: manhattan_spearman
2042
+ value: 87.47315940781353
2043
+ - task:
2044
+ type: STS
2045
+ dataset:
2046
+ type: mteb/sts16-sts
2047
+ name: MTEB STS16
2048
+ config: default
2049
+ split: test
2050
+ revision: 4d8694f8f0e0100860b497b999b3dbed754a0513
2051
+ metrics:
2052
+ - type: cos_sim_pearson
2053
+ value: 83.0230597603627
2054
+ - type: cos_sim_spearman
2055
+ value: 84.93344499318864
2056
+ - type: euclidean_pearson
2057
+ value: 84.23754743431141
2058
+ - type: euclidean_spearman
2059
+ value: 85.09707376597099
2060
+ - type: manhattan_pearson
2061
+ value: 84.04325160987763
2062
+ - type: manhattan_spearman
2063
+ value: 84.89353071339909
2064
+ - task:
2065
+ type: STS
2066
+ dataset:
2067
+ type: mteb/sts17-crosslingual-sts
2068
+ name: MTEB STS17 (en-en)
2069
+ config: en-en
2070
+ split: test
2071
+ revision: af5e6fb845001ecf41f4c1e033ce921939a2a68d
2072
+ metrics:
2073
+ - type: cos_sim_pearson
2074
+ value: 86.75620824563921
2075
+ - type: cos_sim_spearman
2076
+ value: 87.15065513706398
2077
+ - type: euclidean_pearson
2078
+ value: 88.26281533633521
2079
+ - type: euclidean_spearman
2080
+ value: 87.51963738643983
2081
+ - type: manhattan_pearson
2082
+ value: 88.25599267618065
2083
+ - type: manhattan_spearman
2084
+ value: 87.58048736047483
2085
+ - task:
2086
+ type: STS
2087
+ dataset:
2088
+ type: mteb/sts22-crosslingual-sts
2089
+ name: MTEB STS22 (en)
2090
+ config: en
2091
+ split: test
2092
+ revision: 6d1ba47164174a496b7fa5d3569dae26a6813b80
2093
+ metrics:
2094
+ - type: cos_sim_pearson
2095
+ value: 64.74645319195137
2096
+ - type: cos_sim_spearman
2097
+ value: 65.29996325037214
2098
+ - type: euclidean_pearson
2099
+ value: 67.04297794086443
2100
+ - type: euclidean_spearman
2101
+ value: 65.43841726694343
2102
+ - type: manhattan_pearson
2103
+ value: 67.39459955690904
2104
+ - type: manhattan_spearman
2105
+ value: 65.92864704413651
2106
+ - task:
2107
+ type: STS
2108
+ dataset:
2109
+ type: mteb/stsbenchmark-sts
2110
+ name: MTEB STSBenchmark
2111
+ config: default
2112
+ split: test
2113
+ revision: b0fddb56ed78048fa8b90373c8a3cfc37b684831
2114
+ metrics:
2115
+ - type: cos_sim_pearson
2116
+ value: 84.31291020270801
2117
+ - type: cos_sim_spearman
2118
+ value: 85.86473738688068
2119
+ - type: euclidean_pearson
2120
+ value: 85.65537275064152
2121
+ - type: euclidean_spearman
2122
+ value: 86.13087454209642
2123
+ - type: manhattan_pearson
2124
+ value: 85.43946955047609
2125
+ - type: manhattan_spearman
2126
+ value: 85.91568175344916
2127
+ - task:
2128
+ type: Reranking
2129
+ dataset:
2130
+ type: mteb/scidocs-reranking
2131
+ name: MTEB SciDocsRR
2132
+ config: default
2133
+ split: test
2134
+ revision: d3c5e1fc0b855ab6097bf1cda04dd73947d7caab
2135
+ metrics:
2136
+ - type: map
2137
+ value: 85.93798118350695
2138
+ - type: mrr
2139
+ value: 95.93536274908824
2140
+ - task:
2141
+ type: Retrieval
2142
+ dataset:
2143
+ type: scifact
2144
+ name: MTEB SciFact
2145
+ config: default
2146
+ split: test
2147
+ revision: None
2148
+ metrics:
2149
+ - type: map_at_1
2150
+ value: 57.594
2151
+ - type: map_at_10
2152
+ value: 66.81899999999999
2153
+ - type: map_at_100
2154
+ value: 67.368
2155
+ - type: map_at_1000
2156
+ value: 67.4
2157
+ - type: map_at_3
2158
+ value: 64.061
2159
+ - type: map_at_5
2160
+ value: 65.47
2161
+ - type: mrr_at_1
2162
+ value: 60.667
2163
+ - type: mrr_at_10
2164
+ value: 68.219
2165
+ - type: mrr_at_100
2166
+ value: 68.655
2167
+ - type: mrr_at_1000
2168
+ value: 68.684
2169
+ - type: mrr_at_3
2170
+ value: 66.22200000000001
2171
+ - type: mrr_at_5
2172
+ value: 67.289
2173
+ - type: ndcg_at_1
2174
+ value: 60.667
2175
+ - type: ndcg_at_10
2176
+ value: 71.275
2177
+ - type: ndcg_at_100
2178
+ value: 73.642
2179
+ - type: ndcg_at_1000
2180
+ value: 74.373
2181
+ - type: ndcg_at_3
2182
+ value: 66.521
2183
+ - type: ndcg_at_5
2184
+ value: 68.581
2185
+ - type: precision_at_1
2186
+ value: 60.667
2187
+ - type: precision_at_10
2188
+ value: 9.433
2189
+ - type: precision_at_100
2190
+ value: 1.0699999999999998
2191
+ - type: precision_at_1000
2192
+ value: 0.11299999999999999
2193
+ - type: precision_at_3
2194
+ value: 25.556
2195
+ - type: precision_at_5
2196
+ value: 16.8
2197
+ - type: recall_at_1
2198
+ value: 57.594
2199
+ - type: recall_at_10
2200
+ value: 83.622
2201
+ - type: recall_at_100
2202
+ value: 94.167
2203
+ - type: recall_at_1000
2204
+ value: 99.667
2205
+ - type: recall_at_3
2206
+ value: 70.64399999999999
2207
+ - type: recall_at_5
2208
+ value: 75.983
2209
+ - task:
2210
+ type: PairClassification
2211
+ dataset:
2212
+ type: mteb/sprintduplicatequestions-pairclassification
2213
+ name: MTEB SprintDuplicateQuestions
2214
+ config: default
2215
+ split: test
2216
+ revision: d66bd1f72af766a5cc4b0ca5e00c162f89e8cc46
2217
+ metrics:
2218
+ - type: cos_sim_accuracy
2219
+ value: 99.85841584158416
2220
+ - type: cos_sim_ap
2221
+ value: 96.66996142314342
2222
+ - type: cos_sim_f1
2223
+ value: 92.83208020050125
2224
+ - type: cos_sim_precision
2225
+ value: 93.06532663316584
2226
+ - type: cos_sim_recall
2227
+ value: 92.60000000000001
2228
+ - type: dot_accuracy
2229
+ value: 99.85841584158416
2230
+ - type: dot_ap
2231
+ value: 96.6775307676576
2232
+ - type: dot_f1
2233
+ value: 92.69289729177312
2234
+ - type: dot_precision
2235
+ value: 94.77533960292581
2236
+ - type: dot_recall
2237
+ value: 90.7
2238
+ - type: euclidean_accuracy
2239
+ value: 99.86138613861387
2240
+ - type: euclidean_ap
2241
+ value: 96.6338454403108
2242
+ - type: euclidean_f1
2243
+ value: 92.92214357937311
2244
+ - type: euclidean_precision
2245
+ value: 93.96728016359918
2246
+ - type: euclidean_recall
2247
+ value: 91.9
2248
+ - type: manhattan_accuracy
2249
+ value: 99.86237623762376
2250
+ - type: manhattan_ap
2251
+ value: 96.60370449645053
2252
+ - type: manhattan_f1
2253
+ value: 92.91177970423253
2254
+ - type: manhattan_precision
2255
+ value: 94.7970863683663
2256
+ - type: manhattan_recall
2257
+ value: 91.10000000000001
2258
+ - type: max_accuracy
2259
+ value: 99.86237623762376
2260
+ - type: max_ap
2261
+ value: 96.6775307676576
2262
+ - type: max_f1
2263
+ value: 92.92214357937311
2264
+ - task:
2265
+ type: Clustering
2266
+ dataset:
2267
+ type: mteb/stackexchange-clustering
2268
+ name: MTEB StackExchangeClustering
2269
+ config: default
2270
+ split: test
2271
+ revision: 6cbc1f7b2bc0622f2e39d2c77fa502909748c259
2272
+ metrics:
2273
+ - type: v_measure
2274
+ value: 60.77977058695198
2275
+ - task:
2276
+ type: Clustering
2277
+ dataset:
2278
+ type: mteb/stackexchange-clustering-p2p
2279
+ name: MTEB StackExchangeClusteringP2P
2280
+ config: default
2281
+ split: test
2282
+ revision: 815ca46b2622cec33ccafc3735d572c266efdb44
2283
+ metrics:
2284
+ - type: v_measure
2285
+ value: 35.2725272535638
2286
+ - task:
2287
+ type: Reranking
2288
+ dataset:
2289
+ type: mteb/stackoverflowdupquestions-reranking
2290
+ name: MTEB StackOverflowDupQuestions
2291
+ config: default
2292
+ split: test
2293
+ revision: e185fbe320c72810689fc5848eb6114e1ef5ec69
2294
+ metrics:
2295
+ - type: map
2296
+ value: 53.64052466362125
2297
+ - type: mrr
2298
+ value: 54.533067014684654
2299
+ - task:
2300
+ type: Summarization
2301
+ dataset:
2302
+ type: mteb/summeval
2303
+ name: MTEB SummEval
2304
+ config: default
2305
+ split: test
2306
+ revision: cda12ad7615edc362dbf25a00fdd61d3b1eaf93c
2307
+ metrics:
2308
+ - type: cos_sim_pearson
2309
+ value: 30.677624219206578
2310
+ - type: cos_sim_spearman
2311
+ value: 30.121368518123447
2312
+ - type: dot_pearson
2313
+ value: 30.69870088041608
2314
+ - type: dot_spearman
2315
+ value: 29.61284927093751
2316
+ - task:
2317
+ type: Retrieval
2318
+ dataset:
2319
+ type: trec-covid
2320
+ name: MTEB TRECCOVID
2321
+ config: default
2322
+ split: test
2323
+ revision: None
2324
+ metrics:
2325
+ - type: map_at_1
2326
+ value: 0.22
2327
+ - type: map_at_10
2328
+ value: 1.855
2329
+ - type: map_at_100
2330
+ value: 9.885
2331
+ - type: map_at_1000
2332
+ value: 23.416999999999998
2333
+ - type: map_at_3
2334
+ value: 0.637
2335
+ - type: map_at_5
2336
+ value: 1.024
2337
+ - type: mrr_at_1
2338
+ value: 88.0
2339
+ - type: mrr_at_10
2340
+ value: 93.067
2341
+ - type: mrr_at_100
2342
+ value: 93.067
2343
+ - type: mrr_at_1000
2344
+ value: 93.067
2345
+ - type: mrr_at_3
2346
+ value: 92.667
2347
+ - type: mrr_at_5
2348
+ value: 93.067
2349
+ - type: ndcg_at_1
2350
+ value: 82.0
2351
+ - type: ndcg_at_10
2352
+ value: 75.899
2353
+ - type: ndcg_at_100
2354
+ value: 55.115
2355
+ - type: ndcg_at_1000
2356
+ value: 48.368
2357
+ - type: ndcg_at_3
2358
+ value: 79.704
2359
+ - type: ndcg_at_5
2360
+ value: 78.39699999999999
2361
+ - type: precision_at_1
2362
+ value: 88.0
2363
+ - type: precision_at_10
2364
+ value: 79.60000000000001
2365
+ - type: precision_at_100
2366
+ value: 56.06
2367
+ - type: precision_at_1000
2368
+ value: 21.206
2369
+ - type: precision_at_3
2370
+ value: 84.667
2371
+ - type: precision_at_5
2372
+ value: 83.2
2373
+ - type: recall_at_1
2374
+ value: 0.22
2375
+ - type: recall_at_10
2376
+ value: 2.078
2377
+ - type: recall_at_100
2378
+ value: 13.297
2379
+ - type: recall_at_1000
2380
+ value: 44.979
2381
+ - type: recall_at_3
2382
+ value: 0.6689999999999999
2383
+ - type: recall_at_5
2384
+ value: 1.106
2385
+ - task:
2386
+ type: Retrieval
2387
+ dataset:
2388
+ type: webis-touche2020
2389
+ name: MTEB Touche2020
2390
+ config: default
2391
+ split: test
2392
+ revision: None
2393
+ metrics:
2394
+ - type: map_at_1
2395
+ value: 2.258
2396
+ - type: map_at_10
2397
+ value: 10.439
2398
+ - type: map_at_100
2399
+ value: 16.89
2400
+ - type: map_at_1000
2401
+ value: 18.407999999999998
2402
+ - type: map_at_3
2403
+ value: 5.668
2404
+ - type: map_at_5
2405
+ value: 7.718
2406
+ - type: mrr_at_1
2407
+ value: 32.653
2408
+ - type: mrr_at_10
2409
+ value: 51.159
2410
+ - type: mrr_at_100
2411
+ value: 51.714000000000006
2412
+ - type: mrr_at_1000
2413
+ value: 51.714000000000006
2414
+ - type: mrr_at_3
2415
+ value: 47.959
2416
+ - type: mrr_at_5
2417
+ value: 50.407999999999994
2418
+ - type: ndcg_at_1
2419
+ value: 29.592000000000002
2420
+ - type: ndcg_at_10
2421
+ value: 26.037
2422
+ - type: ndcg_at_100
2423
+ value: 37.924
2424
+ - type: ndcg_at_1000
2425
+ value: 49.126999999999995
2426
+ - type: ndcg_at_3
2427
+ value: 30.631999999999998
2428
+ - type: ndcg_at_5
2429
+ value: 28.571
2430
+ - type: precision_at_1
2431
+ value: 32.653
2432
+ - type: precision_at_10
2433
+ value: 22.857
2434
+ - type: precision_at_100
2435
+ value: 7.754999999999999
2436
+ - type: precision_at_1000
2437
+ value: 1.529
2438
+ - type: precision_at_3
2439
+ value: 34.014
2440
+ - type: precision_at_5
2441
+ value: 29.796
2442
+ - type: recall_at_1
2443
+ value: 2.258
2444
+ - type: recall_at_10
2445
+ value: 16.554
2446
+ - type: recall_at_100
2447
+ value: 48.439
2448
+ - type: recall_at_1000
2449
+ value: 82.80499999999999
2450
+ - type: recall_at_3
2451
+ value: 7.283
2452
+ - type: recall_at_5
2453
+ value: 10.732
2454
+ - task:
2455
+ type: Classification
2456
+ dataset:
2457
+ type: mteb/toxic_conversations_50k
2458
+ name: MTEB ToxicConversationsClassification
2459
+ config: default
2460
+ split: test
2461
+ revision: d7c0de2777da35d6aae2200a62c6e0e5af397c4c
2462
+ metrics:
2463
+ - type: accuracy
2464
+ value: 69.8858
2465
+ - type: ap
2466
+ value: 13.835684144362109
2467
+ - type: f1
2468
+ value: 53.803351693244586
2469
+ - task:
2470
+ type: Classification
2471
+ dataset:
2472
+ type: mteb/tweet_sentiment_extraction
2473
+ name: MTEB TweetSentimentExtractionClassification
2474
+ config: default
2475
+ split: test
2476
+ revision: d604517c81ca91fe16a244d1248fc021f9ecee7a
2477
+ metrics:
2478
+ - type: accuracy
2479
+ value: 60.50650820599886
2480
+ - type: f1
2481
+ value: 60.84357825979259
2482
+ - task:
2483
+ type: Clustering
2484
+ dataset:
2485
+ type: mteb/twentynewsgroups-clustering
2486
+ name: MTEB TwentyNewsgroupsClustering
2487
+ config: default
2488
+ split: test
2489
+ revision: 6125ec4e24fa026cec8a478383ee943acfbd5449
2490
+ metrics:
2491
+ - type: v_measure
2492
+ value: 48.52131044852134
2493
+ - task:
2494
+ type: PairClassification
2495
+ dataset:
2496
+ type: mteb/twittersemeval2015-pairclassification
2497
+ name: MTEB TwitterSemEval2015
2498
+ config: default
2499
+ split: test
2500
+ revision: 70970daeab8776df92f5ea462b6173c0b46fd2d1
2501
+ metrics:
2502
+ - type: cos_sim_accuracy
2503
+ value: 85.59337187816654
2504
+ - type: cos_sim_ap
2505
+ value: 73.23925826533437
2506
+ - type: cos_sim_f1
2507
+ value: 67.34693877551021
2508
+ - type: cos_sim_precision
2509
+ value: 62.40432237730752
2510
+ - type: cos_sim_recall
2511
+ value: 73.13984168865434
2512
+ - type: dot_accuracy
2513
+ value: 85.31322644096085
2514
+ - type: dot_ap
2515
+ value: 72.30723963807422
2516
+ - type: dot_f1
2517
+ value: 66.47051612112296
2518
+ - type: dot_precision
2519
+ value: 62.0792305930845
2520
+ - type: dot_recall
2521
+ value: 71.53034300791556
2522
+ - type: euclidean_accuracy
2523
+ value: 85.61125350181797
2524
+ - type: euclidean_ap
2525
+ value: 73.32843720487845
2526
+ - type: euclidean_f1
2527
+ value: 67.36549633745895
2528
+ - type: euclidean_precision
2529
+ value: 64.60755813953489
2530
+ - type: euclidean_recall
2531
+ value: 70.36939313984169
2532
+ - type: manhattan_accuracy
2533
+ value: 85.63509566668654
2534
+ - type: manhattan_ap
2535
+ value: 73.16658488311325
2536
+ - type: manhattan_f1
2537
+ value: 67.20597386434349
2538
+ - type: manhattan_precision
2539
+ value: 63.60424028268551
2540
+ - type: manhattan_recall
2541
+ value: 71.2401055408971
2542
+ - type: max_accuracy
2543
+ value: 85.63509566668654
2544
+ - type: max_ap
2545
+ value: 73.32843720487845
2546
+ - type: max_f1
2547
+ value: 67.36549633745895
2548
+ - task:
2549
+ type: PairClassification
2550
+ dataset:
2551
+ type: mteb/twitterurlcorpus-pairclassification
2552
+ name: MTEB TwitterURLCorpus
2553
+ config: default
2554
+ split: test
2555
+ revision: 8b6510b0b1fa4e4c4f879467980e9be563ec1cdf
2556
+ metrics:
2557
+ - type: cos_sim_accuracy
2558
+ value: 88.33779640625606
2559
+ - type: cos_sim_ap
2560
+ value: 84.83868375898157
2561
+ - type: cos_sim_f1
2562
+ value: 77.16506154017773
2563
+ - type: cos_sim_precision
2564
+ value: 74.62064005753327
2565
+ - type: cos_sim_recall
2566
+ value: 79.88912842623961
2567
+ - type: dot_accuracy
2568
+ value: 88.02732176815307
2569
+ - type: dot_ap
2570
+ value: 83.95089283763002
2571
+ - type: dot_f1
2572
+ value: 76.29635101196631
2573
+ - type: dot_precision
2574
+ value: 73.31771720613288
2575
+ - type: dot_recall
2576
+ value: 79.52725592854944
2577
+ - type: euclidean_accuracy
2578
+ value: 88.44452206310397
2579
+ - type: euclidean_ap
2580
+ value: 84.98384576824827
2581
+ - type: euclidean_f1
2582
+ value: 77.29311047696697
2583
+ - type: euclidean_precision
2584
+ value: 74.51232583065381
2585
+ - type: euclidean_recall
2586
+ value: 80.28949799815214
2587
+ - type: manhattan_accuracy
2588
+ value: 88.47362906042613
2589
+ - type: manhattan_ap
2590
+ value: 84.91421462218432
2591
+ - type: manhattan_f1
2592
+ value: 77.05107637204792
2593
+ - type: manhattan_precision
2594
+ value: 74.74484256243214
2595
+ - type: manhattan_recall
2596
+ value: 79.50415768401602
2597
+ - type: max_accuracy
2598
+ value: 88.47362906042613
2599
+ - type: max_ap
2600
+ value: 84.98384576824827
2601
+ - type: max_f1
2602
+ value: 77.29311047696697
2603
+ license: mit
2604
+ language:
2605
+ - en
2606
+ ---
2607
+
2608
+
2609
+ <h1 align="center">FlagEmbedding</h1>
2610
+
2611
+
2612
+ <h4 align="center">
2613
+ <p>
2614
+ <a href=#model-list>Model List</a> |
2615
+ <a href=#frequently-asked-questions>FAQ</a> |
2616
+ <a href=#usage>Usage</a> |
2617
+ <a href="#evaluation">Evaluation</a> |
2618
+ <a href="#train">Train</a> |
2619
+ <a href="#contact">Contact</a> |
2620
+ <a href="#citation">Citation</a> |
2621
+ <a href="#license">License</a>
2622
+ <p>
2623
+ </h4>
2624
+
2625
+ More details please refer to our Github: [FlagEmbedding](https://github.com/FlagOpen/FlagEmbedding).
2626
+
2627
+ If you are looking for a model that supports more languages, longer texts, and other retrieval methods, you can try using [bge-m3](https://huggingface.co/BAAI/bge-m3).
2628
+
2629
+
2630
+ [English](README.md) | [中文](https://github.com/FlagOpen/FlagEmbedding/blob/master/README_zh.md)
2631
+
2632
+ FlagEmbedding focuses on retrieval-augmented LLMs, consisting of the following projects currently:
2633
+
2634
+ - **Long-Context LLM**: [Activation Beacon](https://github.com/FlagOpen/FlagEmbedding/tree/master/Long_LLM/activation_beacon)
2635
+ - **Fine-tuning of LM** : [LM-Cocktail](https://github.com/FlagOpen/FlagEmbedding/tree/master/LM_Cocktail)
2636
+ - **Dense Retrieval**: [BGE-M3](https://github.com/FlagOpen/FlagEmbedding/tree/master/FlagEmbedding/BGE_M3), [LLM Embedder](https://github.com/FlagOpen/FlagEmbedding/tree/master/FlagEmbedding/llm_embedder), [BGE Embedding](https://github.com/FlagOpen/FlagEmbedding/tree/master/FlagEmbedding/baai_general_embedding)
2637
+ - **Reranker Model**: [BGE Reranker](https://github.com/FlagOpen/FlagEmbedding/tree/master/FlagEmbedding/reranker)
2638
+ - **Benchmark**: [C-MTEB](https://github.com/FlagOpen/FlagEmbedding/tree/master/C_MTEB)
2639
+
2640
+ ## News
2641
+ - 1/30/2024: Release **BGE-M3**, a new member to BGE model series! M3 stands for **M**ulti-linguality (100+ languages), **M**ulti-granularities (input length up to 8192), **M**ulti-Functionality (unification of dense, lexical, multi-vec/colbert retrieval).
2642
+ It is the first embedding model which supports all three retrieval methods, achieving new SOTA on multi-lingual (MIRACL) and cross-lingual (MKQA) benchmarks.
2643
+ [Technical Report](https://github.com/FlagOpen/FlagEmbedding/blob/master/FlagEmbedding/BGE_M3/BGE_M3.pdf) and [Code](https://github.com/FlagOpen/FlagEmbedding/tree/master/FlagEmbedding/BGE_M3). :fire:
2644
+ - 1/9/2024: Release [Activation-Beacon](https://github.com/FlagOpen/FlagEmbedding/tree/master/Long_LLM/activation_beacon), an effective, efficient, compatible, and low-cost (training) method to extend the context length of LLM. [Technical Report](https://arxiv.org/abs/2401.03462) :fire:
2645
+ - 12/24/2023: Release **LLaRA**, a LLaMA-7B based dense retriever, leading to state-of-the-art performances on MS MARCO and BEIR. Model and code will be open-sourced. Please stay tuned. [Technical Report](https://arxiv.org/abs/2312.15503) :fire:
2646
+ - 11/23/2023: Release [LM-Cocktail](https://github.com/FlagOpen/FlagEmbedding/tree/master/LM_Cocktail), a method to maintain general capabilities during fine-tuning by merging multiple language models. [Technical Report](https://arxiv.org/abs/2311.13534) :fire:
2647
+ - 10/12/2023: Release [LLM-Embedder](https://github.com/FlagOpen/FlagEmbedding/tree/master/FlagEmbedding/llm_embedder), a unified embedding model to support diverse retrieval augmentation needs for LLMs. [Technical Report](https://arxiv.org/pdf/2310.07554.pdf)
2648
+ - 09/15/2023: The [technical report](https://arxiv.org/pdf/2309.07597.pdf) of BGE has been released
2649
+ - 09/15/2023: The [massive training data](https://data.baai.ac.cn/details/BAAI-MTP) of BGE has been released
2650
+ - 09/12/2023: New models:
2651
+ - **New reranker model**: release cross-encoder models `BAAI/bge-reranker-base` and `BAAI/bge-reranker-large`, which are more powerful than embedding model. We recommend to use/fine-tune them to re-rank top-k documents returned by embedding models.
2652
+ - **update embedding model**: release `bge-*-v1.5` embedding model to alleviate the issue of the similarity distribution, and enhance its retrieval ability without instruction.
2653
+
2654
+
2655
+ <details>
2656
+ <summary>More</summary>
2657
+ <!-- ### More -->
2658
+
2659
+ - 09/07/2023: Update [fine-tune code](https://github.com/FlagOpen/FlagEmbedding/blob/master/FlagEmbedding/baai_general_embedding/README.md): Add script to mine hard negatives and support adding instruction during fine-tuning.
2660
+ - 08/09/2023: BGE Models are integrated into **Langchain**, you can use it like [this](#using-langchain); C-MTEB **leaderboard** is [available](https://huggingface.co/spaces/mteb/leaderboard).
2661
+ - 08/05/2023: Release base-scale and small-scale models, **best performance among the models of the same size 🤗**
2662
+ - 08/02/2023: Release `bge-large-*`(short for BAAI General Embedding) Models, **rank 1st on MTEB and C-MTEB benchmark!** :tada: :tada:
2663
+ - 08/01/2023: We release the [Chinese Massive Text Embedding Benchmark](https://github.com/FlagOpen/FlagEmbedding/blob/master/C_MTEB) (**C-MTEB**), consisting of 31 test dataset.
2664
+
2665
+ </details>
2666
+
2667
+
2668
+ ## Model List
2669
+
2670
+ `bge` is short for `BAAI general embedding`.
2671
+
2672
+ | Model | Language | | Description | query instruction for retrieval [1] |
2673
+ |:-------------------------------|:--------:| :--------:| :--------:|:--------:|
2674
+ | [BAAI/bge-m3](https://huggingface.co/BAAI/bge-m3) | Multilingual | [Inference](https://github.com/FlagOpen/FlagEmbedding/tree/master/FlagEmbedding/BGE_M3#usage) [Fine-tune](https://github.com/FlagOpen/FlagEmbedding/tree/master/FlagEmbedding/BGE_M3) | Multi-Functionality(dense retrieval, sparse retrieval, multi-vector(colbert)), Multi-Linguality, and Multi-Granularity(8192 tokens) | |
2675
+ | [BAAI/llm-embedder](https://huggingface.co/BAAI/llm-embedder) | English | [Inference](./FlagEmbedding/llm_embedder/README.md) [Fine-tune](./FlagEmbedding/llm_embedder/README.md) | a unified embedding model to support diverse retrieval augmentation needs for LLMs | See [README](./FlagEmbedding/llm_embedder/README.md) |
2676
+ | [BAAI/bge-reranker-large](https://huggingface.co/BAAI/bge-reranker-large) | Chinese and English | [Inference](#usage-for-reranker) [Fine-tune](https://github.com/FlagOpen/FlagEmbedding/tree/master/examples/reranker) | a cross-encoder model which is more accurate but less efficient [2] | |
2677
+ | [BAAI/bge-reranker-base](https://huggingface.co/BAAI/bge-reranker-base) | Chinese and English | [Inference](#usage-for-reranker) [Fine-tune](https://github.com/FlagOpen/FlagEmbedding/tree/master/examples/reranker) | a cross-encoder model which is more accurate but less efficient [2] | |
2678
+ | [BAAI/bge-large-en-v1.5](https://huggingface.co/BAAI/bge-large-en-v1.5) | English | [Inference](#usage-for-embedding-model) [Fine-tune](https://github.com/FlagOpen/FlagEmbedding/tree/master/examples/finetune) | version 1.5 with more reasonable similarity distribution | `Represent this sentence for searching relevant passages: ` |
2679
+ | [BAAI/bge-base-en-v1.5](https://huggingface.co/BAAI/bge-base-en-v1.5) | English | [Inference](#usage-for-embedding-model) [Fine-tune](https://github.com/FlagOpen/FlagEmbedding/tree/master/examples/finetune) | version 1.5 with more reasonable similarity distribution | `Represent this sentence for searching relevant passages: ` |
2680
+ | [BAAI/bge-small-en-v1.5](https://huggingface.co/BAAI/bge-small-en-v1.5) | English | [Inference](#usage-for-embedding-model) [Fine-tune](https://github.com/FlagOpen/FlagEmbedding/tree/master/examples/finetune) | version 1.5 with more reasonable similarity distribution | `Represent this sentence for searching relevant passages: ` |
2681
+ | [BAAI/bge-large-zh-v1.5](https://huggingface.co/BAAI/bge-large-zh-v1.5) | Chinese | [Inference](#usage-for-embedding-model) [Fine-tune](https://github.com/FlagOpen/FlagEmbedding/tree/master/examples/finetune) | version 1.5 with more reasonable similarity distribution | `为这个句子生成表示以用于检索相关文章：` |
2682
+ | [BAAI/bge-base-zh-v1.5](https://huggingface.co/BAAI/bge-base-zh-v1.5) | Chinese | [Inference](#usage-for-embedding-model) [Fine-tune](https://github.com/FlagOpen/FlagEmbedding/tree/master/examples/finetune) | version 1.5 with more reasonable similarity distribution | `为这个句子生成表示以用于检索相关文章：` |
2683
+ | [BAAI/bge-small-zh-v1.5](https://huggingface.co/BAAI/bge-small-zh-v1.5) | Chinese | [Inference](#usage-for-embedding-model) [Fine-tune](https://github.com/FlagOpen/FlagEmbedding/tree/master/examples/finetune) | version 1.5 with more reasonable similarity distribution | `为这个句子生成表示以用于检索相关文章：` |
2684
+ | [BAAI/bge-large-en](https://huggingface.co/BAAI/bge-large-en) | English | [Inference](#usage-for-embedding-model) [Fine-tune](https://github.com/FlagOpen/FlagEmbedding/tree/master/examples/finetune) | :trophy: rank **1st** in [MTEB](https://huggingface.co/spaces/mteb/leaderboard) leaderboard | `Represent this sentence for searching relevant passages: ` |
2685
+ | [BAAI/bge-base-en](https://huggingface.co/BAAI/bge-base-en) | English | [Inference](#usage-for-embedding-model) [Fine-tune](https://github.com/FlagOpen/FlagEmbedding/tree/master/examples/finetune) | a base-scale model but with similar ability to `bge-large-en` | `Represent this sentence for searching relevant passages: ` |
2686
+ | [BAAI/bge-small-en](https://huggingface.co/BAAI/bge-small-en) | English | [Inference](#usage-for-embedding-model) [Fine-tune](https://github.com/FlagOpen/FlagEmbedding/tree/master/examples/finetune) |a small-scale model but with competitive performance | `Represent this sentence for searching relevant passages: ` |
2687
+ | [BAAI/bge-large-zh](https://huggingface.co/BAAI/bge-large-zh) | Chinese | [Inference](#usage-for-embedding-model) [Fine-tune](https://github.com/FlagOpen/FlagEmbedding/tree/master/examples/finetune) | :trophy: rank **1st** in [C-MTEB](https://github.com/FlagOpen/FlagEmbedding/tree/master/C_MTEB) benchmark | `为这个句子生成表示以用于检索相关文章：` |
2688
+ | [BAAI/bge-base-zh](https://huggingface.co/BAAI/bge-base-zh) | Chinese | [Inference](#usage-for-embedding-model) [Fine-tune](https://github.com/FlagOpen/FlagEmbedding/tree/master/examples/finetune) | a base-scale model but with similar ability to `bge-large-zh` | `为这个句子生成表示以用于检索相关文章：` |
2689
+ | [BAAI/bge-small-zh](https://huggingface.co/BAAI/bge-small-zh) | Chinese | [Inference](#usage-for-embedding-model) [Fine-tune](https://github.com/FlagOpen/FlagEmbedding/tree/master/examples/finetune) | a small-scale model but with competitive performance | `为这个句子生成表示以用于检索相关文章：` |
2690
+
2691
+ [1\]: If you need to search the relevant passages to a query, we suggest to add the instruction to the query; in other cases, no instruction is needed, just use the original query directly. In all cases, **no instruction** needs to be added to passages.
2692
+
2693
+ [2\]: Different from embedding model, reranker uses question and document as input and directly output similarity instead of embedding. To balance the accuracy and time cost, cross-encoder is widely used to re-rank top-k documents retrieved by other simple models.
2694
+ For examples, use bge embedding model to retrieve top 100 relevant documents, and then use bge reranker to re-rank the top 100 document to get the final top-3 results.
2695
+
2696
+ All models have been uploaded to Huggingface Hub, and you can see them at https://huggingface.co/BAAI.
2697
+ If you cannot open the Huggingface Hub, you also can download the models at https://model.baai.ac.cn/models .
2698
+
2699
+
2700
+ ## Frequently asked questions
2701
+
2702
+ <details>
2703
+ <summary>1. How to fine-tune bge embedding model?</summary>
2704
+
2705
+ <!-- ### How to fine-tune bge embedding model? -->
2706
+ Following this [example](https://github.com/FlagOpen/FlagEmbedding/tree/master/examples/finetune) to prepare data and fine-tune your model.
2707
+ Some suggestions:
2708
+ - Mine hard negatives following this [example](https://github.com/FlagOpen/FlagEmbedding/tree/master/examples/finetune#hard-negatives), which can improve the retrieval performance.
2709
+ - If you pre-train bge on your data, the pre-trained model cannot be directly used to calculate similarity, and it must be fine-tuned with contrastive learning before computing similarity.
2710
+ - If the accuracy of the fine-tuned model is still not high, it is recommended to use/fine-tune the cross-encoder model (bge-reranker) to re-rank top-k results. Hard negatives also are needed to fine-tune reranker.
2711
+
2712
+
2713
+ </details>
2714
+
2715
+ <details>
2716
+ <summary>2. The similarity score between two dissimilar sentences is higher than 0.5</summary>
2717
+
2718
+ <!-- ### The similarity score between two dissimilar sentences is higher than 0.5 -->
2719
+ **Suggest to use bge v1.5, which alleviates the issue of the similarity distribution.**
2720
+
2721
+ Since we finetune the models by contrastive learning with a temperature of 0.01,
2722
+ the similarity distribution of the current BGE model is about in the interval \[0.6, 1\].
2723
+ So a similarity score greater than 0.5 does not indicate that the two sentences are similar.
2724
+
2725
+ For downstream tasks, such as passage retrieval or semantic similarity,
2726
+ **what matters is the relative order of the scores, not the absolute value.**
2727
+ If you need to filter similar sentences based on a similarity threshold,
2728
+ please select an appropriate similarity threshold based on the similarity distribution on your data (such as 0.8, 0.85, or even 0.9).
2729
+
2730
+ </details>
2731
+
2732
+ <details>
2733
+ <summary>3. When does the query instruction need to be used</summary>
2734
+
2735
+ <!-- ### When does the query instruction need to be used -->
2736
+
2737
+ For the `bge-*-v1.5`, we improve its retrieval ability when not using instruction.
2738
+ No instruction only has a slight degradation in retrieval performance compared with using instruction.
2739
+ So you can generate embedding without instruction in all cases for convenience.
2740
+
2741
+ For a retrieval task that uses short queries to find long related documents,
2742
+ it is recommended to add instructions for these short queries.
2743
+ **The best method to decide whether to add instructions for queries is choosing the setting that achieves better performance on your task.**
2744
+ In all cases, the documents/passages do not need to add the instruction.
2745
+
2746
+ </details>
2747
+
2748
+
2749
+ ## Usage
2750
+
2751
+ ### Usage for Embedding Model
2752
+
2753
+ Here are some examples for using `bge` models with
2754
+ [FlagEmbedding](#using-flagembedding), [Sentence-Transformers](#using-sentence-transformers), [Langchain](#using-langchain), or [Huggingface Transformers](#using-huggingface-transformers).
2755
+
2756
+ #### Using FlagEmbedding
2757
+ ```
2758
+ pip install -U FlagEmbedding
2759
+ ```
2760
+ If it doesn't work for you, you can see [FlagEmbedding](https://github.com/FlagOpen/FlagEmbedding/blob/master/FlagEmbedding/baai_general_embedding/README.md) for more methods to install FlagEmbedding.
2761
+
2762
+ ```python
2763
+ from FlagEmbedding import FlagModel
2764
+ sentences_1 = ["样例数据-1", "样例数据-2"]
2765
+ sentences_2 = ["样例数据-3", "样例数据-4"]
2766
+ model = FlagModel('BAAI/bge-large-zh-v1.5',
2767
+ query_instruction_for_retrieval="为这个句子生成表示以用于检索相关文章：",
2768
+ use_fp16=True) # Setting use_fp16 to True speeds up computation with a slight performance degradation
2769
+ embeddings_1 = model.encode(sentences_1)
2770
+ embeddings_2 = model.encode(sentences_2)
2771
+ similarity = embeddings_1 @ embeddings_2.T
2772
+ print(similarity)
2773
+
2774
+ # for s2p(short query to long passage) retrieval task, suggest to use encode_queries() which will automatically add the instruction to each query
2775
+ # corpus in retrieval task can still use encode() or encode_corpus(), since they don't need instruction
2776
+ queries = ['query_1', 'query_2']
2777
+ passages = ["样例文档-1", "样例文档-2"]
2778
+ q_embeddings = model.encode_queries(queries)
2779
+ p_embeddings = model.encode(passages)
2780
+ scores = q_embeddings @ p_embeddings.T
2781
+ ```
2782
+ For the value of the argument `query_instruction_for_retrieval`, see [Model List](https://github.com/FlagOpen/FlagEmbedding/tree/master#model-list).
2783
+
2784
+ By default, FlagModel will use all available GPUs when encoding. Please set `os.environ["CUDA_VISIBLE_DEVICES"]` to select specific GPUs.
2785
+ You also can set `os.environ["CUDA_VISIBLE_DEVICES"]=""` to make all GPUs unavailable.
2786
+
2787
+
2788
+ #### Using Sentence-Transformers
2789
+
2790
+ You can also use the `bge` models with [sentence-transformers](https://www.SBERT.net):
2791
+
2792
+ ```
2793
+ pip install -U sentence-transformers
2794
+ ```
2795
+ ```python
2796
+ from sentence_transformers import SentenceTransformer
2797
+ sentences_1 = ["样例数据-1", "样例数据-2"]
2798
+ sentences_2 = ["样例数据-3", "样例数据-4"]
2799
+ model = SentenceTransformer('BAAI/bge-large-zh-v1.5')
2800
+ embeddings_1 = model.encode(sentences_1, normalize_embeddings=True)
2801
+ embeddings_2 = model.encode(sentences_2, normalize_embeddings=True)
2802
+ similarity = embeddings_1 @ embeddings_2.T
2803
+ print(similarity)
2804
+ ```
2805
+ For s2p(short query to long passage) retrieval task,
2806
+ each short query should start with an instruction (instructions see [Model List](https://github.com/FlagOpen/FlagEmbedding/tree/master#model-list)).
2807
+ But the instruction is not needed for passages.
2808
+ ```python
2809
+ from sentence_transformers import SentenceTransformer
2810
+ queries = ['query_1', 'query_2']
2811
+ passages = ["样例文档-1", "样例文档-2"]
2812
+ instruction = "为这个句子生成表示以用于检索相关文章："
2813
+
2814
+ model = SentenceTransformer('BAAI/bge-large-zh-v1.5')
2815
+ q_embeddings = model.encode([instruction+q for q in queries], normalize_embeddings=True)
2816
+ p_embeddings = model.encode(passages, normalize_embeddings=True)
2817
+ scores = q_embeddings @ p_embeddings.T
2818
+ ```
2819
+
2820
+ #### Using Langchain
2821
+
2822
+ You can use `bge` in langchain like this:
2823
+ ```python
2824
+ from langchain.embeddings import HuggingFaceBgeEmbeddings
2825
+ model_name = "BAAI/bge-large-en-v1.5"
2826
+ model_kwargs = {'device': 'cuda'}
2827
+ encode_kwargs = {'normalize_embeddings': True} # set True to compute cosine similarity
2828
+ model = HuggingFaceBgeEmbeddings(
2829
+ model_name=model_name,
2830
+ model_kwargs=model_kwargs,
2831
+ encode_kwargs=encode_kwargs,
2832
+ query_instruction="为这个句子生成表示以用于检索相关文章："
2833
+ )
2834
+ model.query_instruction = "为这个句子生成表示以用于检索相关文章："
2835
+ ```
2836
+
2837
+
2838
+ #### Using HuggingFace Transformers
2839
+
2840
+ With the transformers package, you can use the model like this: First, you pass your input through the transformer model, then you select the last hidden state of the first token (i.e., [CLS]) as the sentence embedding.
2841
+
2842
+ ```python
2843
+ from transformers import AutoTokenizer, AutoModel
2844
+ import torch
2845
+ # Sentences we want sentence embeddings for
2846
+ sentences = ["样例数据-1", "样例数据-2"]
2847
+
2848
+ # Load model from HuggingFace Hub
2849
+ tokenizer = AutoTokenizer.from_pretrained('BAAI/bge-large-zh-v1.5')
2850
+ model = AutoModel.from_pretrained('BAAI/bge-large-zh-v1.5')
2851
+ model.eval()
2852
+
2853
+ # Tokenize sentences
2854
+ encoded_input = tokenizer(sentences, padding=True, truncation=True, return_tensors='pt')
2855
+ # for s2p(short query to long passage) retrieval task, add an instruction to query (not add instruction for passages)
2856
+ # encoded_input = tokenizer([instruction + q for q in queries], padding=True, truncation=True, return_tensors='pt')
2857
+
2858
+ # Compute token embeddings
2859
+ with torch.no_grad():
2860
+ model_output = model(**encoded_input)
2861
+ # Perform pooling. In this case, cls pooling.
2862
+ sentence_embeddings = model_output[0][:, 0]
2863
+ # normalize embeddings
2864
+ sentence_embeddings = torch.nn.functional.normalize(sentence_embeddings, p=2, dim=1)
2865
+ print("Sentence embeddings:", sentence_embeddings)
2866
+ ```
2867
+
2868
+ ### Usage for Reranker
2869
+
2870
+ Different from embedding model, reranker uses question and document as input and directly output similarity instead of embedding.
2871
+ You can get a relevance score by inputting query and passage to the reranker.
2872
+ The reranker is optimized based cross-entropy loss, so the relevance score is not bounded to a specific range.
2873
+
2874
+
2875
+ #### Using FlagEmbedding
2876
+ ```
2877
+ pip install -U FlagEmbedding
2878
+ ```
2879
+
2880
+ Get relevance scores (higher scores indicate more relevance):
2881
+ ```python
2882
+ from FlagEmbedding import FlagReranker
2883
+ reranker = FlagReranker('BAAI/bge-reranker-large', use_fp16=True) # Setting use_fp16 to True speeds up computation with a slight performance degradation
2884
+
2885
+ score = reranker.compute_score(['query', 'passage'])
2886
+ print(score)
2887
+
2888
+ scores = reranker.compute_score([['what is panda?', 'hi'], ['what is panda?', 'The giant panda (Ailuropoda melanoleuca), sometimes called a panda bear or simply panda, is a bear species endemic to China.']])
2889
+ print(scores)
2890
+ ```
2891
+
2892
+
2893
+ #### Using Huggingface transformers
2894
+
2895
+ ```python
2896
+ import torch
2897
+ from transformers import AutoModelForSequenceClassification, AutoTokenizer
2898
+
2899
+ tokenizer = AutoTokenizer.from_pretrained('BAAI/bge-reranker-large')
2900
+ model = AutoModelForSequenceClassification.from_pretrained('BAAI/bge-reranker-large')
2901
+ model.eval()
2902
+
2903
+ pairs = [['what is panda?', 'hi'], ['what is panda?', 'The giant panda (Ailuropoda melanoleuca), sometimes called a panda bear or simply panda, is a bear species endemic to China.']]
2904
+ with torch.no_grad():
2905
+ inputs = tokenizer(pairs, padding=True, truncation=True, return_tensors='pt', max_length=512)
2906
+ scores = model(**inputs, return_dict=True).logits.view(-1, ).float()
2907
+ print(scores)
2908
+ ```
2909
+
2910
+ #### Usage of the ONNX files
2911
+
2912
+ ```python
2913
+ from optimum.onnxruntime import ORTModelForFeatureExtraction # type: ignore
2914
+
2915
+ import torch
2916
+ from transformers import AutoModel, AutoTokenizer
2917
+
2918
+ tokenizer = AutoTokenizer.from_pretrained('BAAI/bge-small-en-v1.5')
2919
+ model = AutoModel.from_pretrained('BAAI/bge-small-en-v1.5')
2920
+ model_ort = ORTModelForFeatureExtraction.from_pretrained('BAAI/bge-small-en-v1.5', file_name="onnx/model.onnx")
2921
+
2922
+ # Sentences we want sentence embeddings for
2923
+ sentences = ["样例数据-1", "样例数据-2"]
2924
+
2925
+ # Tokenize sentences
2926
+ encoded_input = tokenizer(sentences, padding=True, truncation=True, return_tensors='pt')
2927
+ # for s2p(short query to long passage) retrieval task, add an instruction to query (not add instruction for passages)
2928
+ # encoded_input = tokenizer([instruction + q for q in queries], padding=True, truncation=True, return_tensors='pt')
2929
+
2930
+ model_output_ort = model_ort(**encoded_input)
2931
+ # Compute token embeddings
2932
+ with torch.no_grad():
2933
+ model_output = model(**encoded_input)
2934
+
2935
+ # model_output and model_output_ort are identical
2936
+
2937
+ ```
2938
+
2939
+ #### Usage via infinity
2940
+ Its also possible to deploy the onnx files with the [infinity_emb](https://github.com/michaelfeil/infinity) pip package.
2941
+ Recommended is `device="cuda", engine="torch"` with flash attention on gpu, and `device="cpu", engine="optimum"` for onnx inference.
2942
+
2943
+ ```python
2944
+ import asyncio
2945
+ from infinity_emb import AsyncEmbeddingEngine, EngineArgs
2946
+
2947
+ sentences = ["Embed this is sentence via Infinity.", "Paris is in France."]
2948
+ engine = AsyncEmbeddingEngine.from_args(
2949
+ EngineArgs(model_name_or_path = "BAAI/bge-small-en-v1.5", device="cpu", engine="optimum" # or engine="torch"
2950
+ ))
2951
+
2952
+ async def main():
2953
+ async with engine:
2954
+ embeddings, usage = await engine.embed(sentences=sentences)
2955
+ asyncio.run(main())
2956
+ ```
2957
+
2958
+
2959
+ ## Evaluation
2960
+
2961
+ `baai-general-embedding` models achieve **state-of-the-art performance on both MTEB and C-MTEB leaderboard!**
2962
+ For more details and evaluation tools see our [scripts](https://github.com/FlagOpen/FlagEmbedding/blob/master/C_MTEB/README.md).
2963
+
2964
+ - **MTEB**:
2965
+
2966
+ | Model Name | Dimension | Sequence Length | Average (56) | Retrieval (15) |Clustering (11) | Pair Classification (3) | Reranking (4) | STS (10) | Summarization (1) | Classification (12) |
2967
+ |:----:|:---:|:---:|:---:|:---:|:---:|:---:|:---:|:---:|:---:|:---:|
2968
+ | [BAAI/bge-large-en-v1.5](https://huggingface.co/BAAI/bge-large-en-v1.5) | 1024 | 512 | **64.23** | **54.29** | 46.08 | 87.12 | 60.03 | 83.11 | 31.61 | 75.97 |
2969
+ | [BAAI/bge-base-en-v1.5](https://huggingface.co/BAAI/bge-base-en-v1.5) | 768 | 512 | 63.55 | 53.25 | 45.77 | 86.55 | 58.86 | 82.4 | 31.07 | 75.53 |
2970
+ | [BAAI/bge-small-en-v1.5](https://huggingface.co/BAAI/bge-small-en-v1.5) | 384 | 512 | 62.17 |51.68 | 43.82 | 84.92 | 58.36 | 81.59 | 30.12 | 74.14 |
2971
+ | [bge-large-en](https://huggingface.co/BAAI/bge-large-en) | 1024 | 512 | 63.98 | 53.9 | 46.98 | 85.8 | 59.48 | 81.56 | 32.06 | 76.21 |
2972
+ | [bge-base-en](https://huggingface.co/BAAI/bge-base-en) | 768 | 512 | 63.36 | 53.0 | 46.32 | 85.86 | 58.7 | 81.84 | 29.27 | 75.27 |
2973
+ | [gte-large](https://huggingface.co/thenlper/gte-large) | 1024 | 512 | 63.13 | 52.22 | 46.84 | 85.00 | 59.13 | 83.35 | 31.66 | 73.33 |
2974
+ | [gte-base](https://huggingface.co/thenlper/gte-base) | 768 | 512 | 62.39 | 51.14 | 46.2 | 84.57 | 58.61 | 82.3 | 31.17 | 73.01 |
2975
+ | [e5-large-v2](https://huggingface.co/intfloat/e5-large-v2) | 1024| 512 | 62.25 | 50.56 | 44.49 | 86.03 | 56.61 | 82.05 | 30.19 | 75.24 |
2976
+ | [bge-small-en](https://huggingface.co/BAAI/bge-small-en) | 384 | 512 | 62.11 | 51.82 | 44.31 | 83.78 | 57.97 | 80.72 | 30.53 | 74.37 |
2977
+ | [instructor-xl](https://huggingface.co/hkunlp/instructor-xl) | 768 | 512 | 61.79 | 49.26 | 44.74 | 86.62 | 57.29 | 83.06 | 32.32 | 61.79 |
2978
+ | [e5-base-v2](https://huggingface.co/intfloat/e5-base-v2) | 768 | 512 | 61.5 | 50.29 | 43.80 | 85.73 | 55.91 | 81.05 | 30.28 | 73.84 |
2979
+ | [gte-small](https://huggingface.co/thenlper/gte-small) | 384 | 512 | 61.36 | 49.46 | 44.89 | 83.54 | 57.7 | 82.07 | 30.42 | 72.31 |
2980
+ | [text-embedding-ada-002](https://platform.openai.com/docs/guides/embeddings) | 1536 | 8192 | 60.99 | 49.25 | 45.9 | 84.89 | 56.32 | 80.97 | 30.8 | 70.93 |
2981
+ | [e5-small-v2](https://huggingface.co/intfloat/e5-base-v2) | 384 | 512 | 59.93 | 49.04 | 39.92 | 84.67 | 54.32 | 80.39 | 31.16 | 72.94 |
2982
+ | [sentence-t5-xxl](https://huggingface.co/sentence-transformers/sentence-t5-xxl) | 768 | 512 | 59.51 | 42.24 | 43.72 | 85.06 | 56.42 | 82.63 | 30.08 | 73.42 |
2983
+ | [all-mpnet-base-v2](https://huggingface.co/sentence-transformers/all-mpnet-base-v2) | 768 | 514 | 57.78 | 43.81 | 43.69 | 83.04 | 59.36 | 80.28 | 27.49 | 65.07 |
2984
+ | [sgpt-bloom-7b1-msmarco](https://huggingface.co/bigscience/sgpt-bloom-7b1-msmarco) | 4096 | 2048 | 57.59 | 48.22 | 38.93 | 81.9 | 55.65 | 77.74 | 33.6 | 66.19 |
2985
+
2986
+
2987
+
2988
+ - **C-MTEB**:
2989
+ We create the benchmark C-MTEB for Chinese text embedding which consists of 31 datasets from 6 tasks.
2990
+ Please refer to [C_MTEB](https://github.com/FlagOpen/FlagEmbedding/blob/master/C_MTEB/README.md) for a detailed introduction.
2991
+
2992
+ | Model | Embedding dimension | Avg | Retrieval | STS | PairClassification | Classification | Reranking | Clustering |
2993
+ |:-------------------------------|:--------:|:--------:|:--------:|:--------:|:--------:|:--------:|:--------:|:--------:|
2994
+ | [**BAAI/bge-large-zh-v1.5**](https://huggingface.co/BAAI/bge-large-zh-v1.5) | 1024 | **64.53** | 70.46 | 56.25 | 81.6 | 69.13 | 65.84 | 48.99 |
2995
+ | [BAAI/bge-base-zh-v1.5](https://huggingface.co/BAAI/bge-base-zh-v1.5) | 768 | 63.13 | 69.49 | 53.72 | 79.75 | 68.07 | 65.39 | 47.53 |
2996
+ | [BAAI/bge-small-zh-v1.5](https://huggingface.co/BAAI/bge-small-zh-v1.5) | 512 | 57.82 | 61.77 | 49.11 | 70.41 | 63.96 | 60.92 | 44.18 |
2997
+ | [BAAI/bge-large-zh](https://huggingface.co/BAAI/bge-large-zh) | 1024 | 64.20 | 71.53 | 54.98 | 78.94 | 68.32 | 65.11 | 48.39 |
2998
+ | [bge-large-zh-noinstruct](https://huggingface.co/BAAI/bge-large-zh-noinstruct) | 1024 | 63.53 | 70.55 | 53 | 76.77 | 68.58 | 64.91 | 50.01 |
2999
+ | [BAAI/bge-base-zh](https://huggingface.co/BAAI/bge-base-zh) | 768 | 62.96 | 69.53 | 54.12 | 77.5 | 67.07 | 64.91 | 47.63 |
3000
+ | [multilingual-e5-large](https://huggingface.co/intfloat/multilingual-e5-large) | 1024 | 58.79 | 63.66 | 48.44 | 69.89 | 67.34 | 56.00 | 48.23 |
3001
+ | [BAAI/bge-small-zh](https://huggingface.co/BAAI/bge-small-zh) | 512 | 58.27 | 63.07 | 49.45 | 70.35 | 63.64 | 61.48 | 45.09 |
3002
+ | [m3e-base](https://huggingface.co/moka-ai/m3e-base) | 768 | 57.10 | 56.91 | 50.47 | 63.99 | 67.52 | 59.34 | 47.68 |
3003
+ | [m3e-large](https://huggingface.co/moka-ai/m3e-large) | 1024 | 57.05 | 54.75 | 50.42 | 64.3 | 68.2 | 59.66 | 48.88 |
3004
+ | [multilingual-e5-base](https://huggingface.co/intfloat/multilingual-e5-base) | 768 | 55.48 | 61.63 | 46.49 | 67.07 | 65.35 | 54.35 | 40.68 |
3005
+ | [multilingual-e5-small](https://huggingface.co/intfloat/multilingual-e5-small) | 384 | 55.38 | 59.95 | 45.27 | 66.45 | 65.85 | 53.86 | 45.26 |
3006
+ | [text-embedding-ada-002(OpenAI)](https://platform.openai.com/docs/guides/embeddings/what-are-embeddings) | 1536 | 53.02 | 52.0 | 43.35 | 69.56 | 64.31 | 54.28 | 45.68 |
3007
+ | [luotuo](https://huggingface.co/silk-road/luotuo-bert-medium) | 1024 | 49.37 | 44.4 | 42.78 | 66.62 | 61 | 49.25 | 44.39 |
3008
+ | [text2vec-base](https://huggingface.co/shibing624/text2vec-base-chinese) | 768 | 47.63 | 38.79 | 43.41 | 67.41 | 62.19 | 49.45 | 37.66 |
3009
+ | [text2vec-large](https://huggingface.co/GanymedeNil/text2vec-large-chinese) | 1024 | 47.36 | 41.94 | 44.97 | 70.86 | 60.66 | 49.16 | 30.02 |
3010
+
3011
+
3012
+ - **Reranking**:
3013
+ See [C_MTEB](https://github.com/FlagOpen/FlagEmbedding/blob/master/C_MTEB/) for evaluation script.
3014
+
3015
+ | Model | T2Reranking | T2RerankingZh2En\* | T2RerankingEn2Zh\* | MMarcoReranking | CMedQAv1 | CMedQAv2 | Avg |
3016
+ |:-------------------------------|:--------:|:--------:|:--------:|:--------:|:--------:|:--------:|:--------:|
3017
+ | text2vec-base-multilingual | 64.66 | 62.94 | 62.51 | 14.37 | 48.46 | 48.6 | 50.26 |
3018
+ | multilingual-e5-small | 65.62 | 60.94 | 56.41 | 29.91 | 67.26 | 66.54 | 57.78 |
3019
+ | multilingual-e5-large | 64.55 | 61.61 | 54.28 | 28.6 | 67.42 | 67.92 | 57.4 |
3020
+ | multilingual-e5-base | 64.21 | 62.13 | 54.68 | 29.5 | 66.23 | 66.98 | 57.29 |
3021
+ | m3e-base | 66.03 | 62.74 | 56.07 | 17.51 | 77.05 | 76.76 | 59.36 |
3022
+ | m3e-large | 66.13 | 62.72 | 56.1 | 16.46 | 77.76 | 78.27 | 59.57 |
3023
+ | bge-base-zh-v1.5 | 66.49 | 63.25 | 57.02 | 29.74 | 80.47 | 84.88 | 63.64 |
3024
+ | bge-large-zh-v1.5 | 65.74 | 63.39 | 57.03 | 28.74 | 83.45 | 85.44 | 63.97 |
3025
+ | [BAAI/bge-reranker-base](https://huggingface.co/BAAI/bge-reranker-base) | 67.28 | 63.95 | 60.45 | 35.46 | 81.26 | 84.1 | 65.42 |
3026
+ | [BAAI/bge-reranker-large](https://huggingface.co/BAAI/bge-reranker-large) | 67.6 | 64.03 | 61.44 | 37.16 | 82.15 | 84.18 | 66.09 |
3027
+
3028
+ \* : T2RerankingZh2En and T2RerankingEn2Zh are cross-language retrieval tasks
3029
+
3030
+ ## Train
3031
+
3032
+ ### BAAI Embedding
3033
+
3034
+ We pre-train the models using [retromae](https://github.com/staoxiao/RetroMAE) and train them on large-scale pairs data using contrastive learning.
3035
+ **You can fine-tune the embedding model on your data following our [examples](https://github.com/FlagOpen/FlagEmbedding/tree/master/examples/finetune).**
3036
+ We also provide a [pre-train example](https://github.com/FlagOpen/FlagEmbedding/tree/master/examples/pretrain).
3037
+ Note that the goal of pre-training is to reconstruct the text, and the pre-trained model cannot be used for similarity calculation directly, it needs to be fine-tuned.
3038
+ More training details for bge see [baai_general_embedding](https://github.com/FlagOpen/FlagEmbedding/blob/master/FlagEmbedding/baai_general_embedding/README.md).
3039
+
3040
+
3041
+
3042
+ ### BGE Reranker
3043
+
3044
+ Cross-encoder will perform full-attention over the input pair,
3045
+ which is more accurate than embedding model (i.e., bi-encoder) but more time-consuming than embedding model.
3046
+ Therefore, it can be used to re-rank the top-k documents returned by embedding model.
3047
+ We train the cross-encoder on a multilingual pair data,
3048
+ The data format is the same as embedding model, so you can fine-tune it easily following our [example](https://github.com/FlagOpen/FlagEmbedding/tree/master/examples/reranker).
3049
+ More details please refer to [./FlagEmbedding/reranker/README.md](https://github.com/FlagOpen/FlagEmbedding/tree/master/FlagEmbedding/reranker)
3050
+
3051
+
3052
+ ## Contact
3053
+ If you have any question or suggestion related to this project, feel free to open an issue or pull request.
3054
+ You also can email Shitao Xiao(stxiao@baai.ac.cn) and Zheng Liu(liuzheng@baai.ac.cn).
3055
+
3056
+
3057
+ ## Citation
3058
+
3059
+ If you find this repository useful, please consider giving a star :star: and citation
3060
+
3061
+ ```
3062
+ @misc{bge_embedding,
3063
+ title={C-Pack: Packaged Resources To Advance General Chinese Embedding},
3064
+ author={Shitao Xiao and Zheng Liu and Peitian Zhang and Niklas Muennighoff},
3065
+ year={2023},
3066
+ eprint={2309.07597},
3067
+ archivePrefix={arXiv},
3068
+ primaryClass={cs.CL}
3069
+ }
3070
+ ```
3071
+
3072
+ ## License
3073
+ FlagEmbedding is licensed under the [MIT License](https://github.com/FlagOpen/FlagEmbedding/blob/master/LICENSE). The released models can be used for commercial purposes free of charge.
3074
+
embedder/config.json ADDED
@@ -0,0 +1,35 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_cross_attention": false,
3
+ "architectures": [
4
+ "BertModel"
5
+ ],
6
+ "attention_probs_dropout_prob": 0.1,
7
+ "bos_token_id": null,
8
+ "classifier_dropout": null,
9
+ "dtype": "float32",
10
+ "eos_token_id": null,
11
+ "hidden_act": "gelu",
12
+ "hidden_dropout_prob": 0.1,
13
+ "hidden_size": 384,
14
+ "id2label": {
15
+ "0": "LABEL_0"
16
+ },
17
+ "initializer_range": 0.02,
18
+ "intermediate_size": 1536,
19
+ "is_decoder": false,
20
+ "label2id": {
21
+ "LABEL_0": 0
22
+ },
23
+ "layer_norm_eps": 1e-12,
24
+ "max_position_embeddings": 512,
25
+ "model_type": "bert",
26
+ "num_attention_heads": 12,
27
+ "num_hidden_layers": 12,
28
+ "pad_token_id": 0,
29
+ "position_embedding_type": "absolute",
30
+ "tie_word_embeddings": true,
31
+ "transformers_version": "5.12.1",
32
+ "type_vocab_size": 2,
33
+ "use_cache": true,
34
+ "vocab_size": 30522
35
+ }
embedder/config_sentence_transformers.json ADDED
@@ -0,0 +1,14 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "__version__": {
3
+ "pytorch": "2.12.1",
4
+ "sentence_transformers": "5.6.0",
5
+ "transformers": "5.12.1"
6
+ },
7
+ "default_prompt_name": null,
8
+ "model_type": "SentenceTransformer",
9
+ "prompts": {
10
+ "document": "",
11
+ "query": ""
12
+ },
13
+ "similarity_fn_name": "cosine"
14
+ }
embedder/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ea1d11a3f23d14fe09fc1826fc7944e89c09a634d2217d57a21dd136805ee3e8
3
+ size 133462128
embedder/modules.json ADDED
@@ -0,0 +1,20 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ [
2
+ {
3
+ "idx": 0,
4
+ "name": "0",
5
+ "path": "",
6
+ "type": "sentence_transformers.base.modules.transformer.Transformer"
7
+ },
8
+ {
9
+ "idx": 1,
10
+ "name": "1",
11
+ "path": "1_Pooling",
12
+ "type": "sentence_transformers.sentence_transformer.modules.pooling.Pooling"
13
+ },
14
+ {
15
+ "idx": 2,
16
+ "name": "2",
17
+ "path": "2_Normalize",
18
+ "type": "sentence_transformers.sentence_transformer.modules.normalize.Normalize"
19
+ }
20
+ ]
embedder/sentence_bert_config.json ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "transformer_task": "feature-extraction",
3
+ "modality_config": {
4
+ "text": {
5
+ "method": "forward",
6
+ "method_output_name": "last_hidden_state"
7
+ }
8
+ },
9
+ "module_output_name": "token_embeddings"
10
+ }
embedder/tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
embedder/tokenizer_config.json ADDED
@@ -0,0 +1,18 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "backend": "tokenizers",
3
+ "clean_up_tokenization_spaces": true,
4
+ "cls_token": "[CLS]",
5
+ "do_basic_tokenize": true,
6
+ "do_lower_case": true,
7
+ "is_local": false,
8
+ "local_files_only": false,
9
+ "mask_token": "[MASK]",
10
+ "model_max_length": 512,
11
+ "never_split": null,
12
+ "pad_token": "[PAD]",
13
+ "sep_token": "[SEP]",
14
+ "strip_accents": null,
15
+ "tokenize_chinese_chars": true,
16
+ "tokenizer_class": "BertTokenizer",
17
+ "unk_token": "[UNK]"
18
+ }
modeling_baa.py ADDED
@@ -0,0 +1,83 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ """baa.ai unified Embedding+Reranker loader (generic).
2
+
3
+ A single artifact that does both halves of RAG retrieval — bi-encoder embedding AND cross-encoder reranking —
4
+ over ONE shared word-embedding table. The reranker's word-embedding matrix is stored only once (in the
5
+ embedder) and injected at load, so the packaged model is smaller than shipping the two components separately,
6
+ at no measured quality cost.
7
+
8
+ Works for BERT-based and XLM-RoBERTa-based stacks alike: the reranker's encoder submodule is resolved
9
+ generically via `reranker.base_model` (so `.bert` / `.roberta` are both handled). Optional per-model query/doc
10
+ prompts are read from config.json (e.g. arctic uses "query: ").
11
+
12
+ Usage:
13
+ from modeling_baa import BaaEmbeddingReranker
14
+ m = BaaEmbeddingReranker("path/to/model-dir")
15
+ qv = m.embed(["what is a cross-encoder?"], is_query=True) # normalized bi-encoder vectors
16
+ dv = m.embed(["a cross-encoder scores a (query, doc) pair jointly"])
17
+ ranked = m.rerank("what is a cross-encoder?", ["doc A ...", "doc B ..."]) # [(doc, score), ...] desc
18
+ """
19
+ import os, json, torch
20
+ from safetensors.torch import load_file
21
+ from sentence_transformers import SentenceTransformer
22
+ from transformers import AutoConfig, AutoModelForSequenceClassification, AutoTokenizer
23
+
24
+
25
+ class BaaEmbeddingReranker:
26
+ def __init__(self, path=None, device=None):
27
+ path = path or os.path.dirname(os.path.abspath(__file__))
28
+ self.device = device or ("mps" if torch.backends.mps.is_available()
29
+ else ("cuda" if torch.cuda.is_available() else "cpu"))
30
+ cfg = {}
31
+ cfg_path = os.path.join(path, "config.json")
32
+ if os.path.exists(cfg_path):
33
+ cfg = json.load(open(cfg_path))
34
+ self.q_prompt = cfg.get("embed_query_prompt", "") or ""
35
+ self.d_prompt = cfg.get("embed_doc_prompt", "") or ""
36
+ trc = bool(cfg.get("trust_remote_code", False))
37
+ emb_dir, rr_dir = os.path.join(path, "embedder"), os.path.join(path, "reranker")
38
+
39
+ # embedder = bi-encoder; holds the canonical shared word-embedding table
40
+ self.embedder = SentenceTransformer(emb_dir, device=self.device, trust_remote_code=trc)
41
+ shared_wemb = self.embedder[0].auto_model.embeddings.word_embeddings.weight.data
42
+
43
+ # reranker = cross-encoder seq-classifier; word-embedding stripped on disk -> injected from shared table
44
+ rr_cfg = AutoConfig.from_pretrained(rr_dir, trust_remote_code=trc)
45
+ self.reranker = AutoModelForSequenceClassification.from_config(rr_cfg, trust_remote_code=trc)
46
+ self.reranker.load_state_dict(load_file(os.path.join(rr_dir, "model.safetensors")), strict=False)
47
+ # resolve the encoder submodule generically (.bert for BERT, .roberta for XLM-R, ...)
48
+ base = self.reranker.base_model
49
+ base.embeddings.word_embeddings.weight.data = shared_wemb.to(self.reranker.dtype).clone()
50
+ self.reranker.to(self.device).eval()
51
+ self.rr_tok = AutoTokenizer.from_pretrained(rr_dir, trust_remote_code=trc)
52
+
53
+ def embed(self, texts, is_query=False, batch_size=32):
54
+ """Return L2-normalized bi-encoder vectors. Applies the model's query/doc prompt if configured."""
55
+ prompt = self.q_prompt if is_query else self.d_prompt
56
+ texts = [prompt + t for t in texts] if prompt else list(texts)
57
+ return self.embedder.encode(texts, normalize_embeddings=True,
58
+ batch_size=batch_size, show_progress_bar=False)
59
+
60
+ @torch.no_grad()
61
+ def rerank(self, query, docs, top_k=None, batch_size=32):
62
+ """Cross-encoder relevance scores for (query, doc) pairs; returns [(doc, score), ...] sorted desc."""
63
+ scores = []
64
+ for i in range(0, len(docs), batch_size):
65
+ enc = self.rr_tok([(query, d[:2000]) for d in docs[i:i + batch_size]], padding=True,
66
+ truncation=True, max_length=512, return_tensors="pt").to(self.device)
67
+ scores.extend(self.reranker(**enc).logits[:, 0].float().cpu().tolist())
68
+ order = sorted(range(len(docs)), key=lambda j: -scores[j])
69
+ if top_k:
70
+ order = order[:top_k]
71
+ return [(docs[j], scores[j]) for j in order]
72
+
73
+
74
+ if __name__ == "__main__":
75
+ import numpy as np
76
+ m = BaaEmbeddingReranker()
77
+ q = "How does a cross-encoder reranker work?"
78
+ docs = ["A cross-encoder jointly encodes the query and document to score relevance.",
79
+ "The mitochondria is the powerhouse of the cell.",
80
+ "Bi-encoders embed query and document separately for fast retrieval."]
81
+ qv = m.embed([q], is_query=True)[0]; dv = m.embed(docs)
82
+ print("embed cos:", [round(float(np.dot(qv, d)), 3) for d in dv])
83
+ print("rerank :", [(round(s, 2), d[:45]) for d, s in m.rerank(q, docs)])
reranker/config.json ADDED
@@ -0,0 +1,37 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_cross_attention": false,
3
+ "architectures": [
4
+ "BertForSequenceClassification"
5
+ ],
6
+ "attention_probs_dropout_prob": 0.1,
7
+ "bos_token_id": null,
8
+ "classifier_dropout": null,
9
+ "dtype": "float32",
10
+ "eos_token_id": null,
11
+ "gradient_checkpointing": false,
12
+ "hidden_act": "gelu",
13
+ "hidden_dropout_prob": 0.1,
14
+ "hidden_size": 384,
15
+ "id2label": {
16
+ "0": "LABEL_0"
17
+ },
18
+ "initializer_range": 0.02,
19
+ "intermediate_size": 1536,
20
+ "is_decoder": false,
21
+ "label2id": {
22
+ "LABEL_0": 0
23
+ },
24
+ "layer_norm_eps": 1e-12,
25
+ "max_position_embeddings": 512,
26
+ "model_type": "bert",
27
+ "num_attention_heads": 12,
28
+ "num_hidden_layers": 6,
29
+ "pad_token_id": 0,
30
+ "position_embedding_type": "absolute",
31
+ "sbert_ce_default_activation_function": "torch.nn.modules.linear.Identity",
32
+ "tie_word_embeddings": true,
33
+ "transformers_version": "5.12.1",
34
+ "type_vocab_size": 2,
35
+ "use_cache": true,
36
+ "vocab_size": 30522
37
+ }
reranker/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dd6a2f5080f90f29151244d573042eb42fbb476b388b53dcb6c05429e243bf2d
3
+ size 43984420
reranker/tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
reranker/tokenizer_config.json ADDED
@@ -0,0 +1,18 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "backend": "tokenizers",
3
+ "clean_up_tokenization_spaces": true,
4
+ "cls_token": "[CLS]",
5
+ "do_basic_tokenize": true,
6
+ "do_lower_case": true,
7
+ "is_local": false,
8
+ "local_files_only": false,
9
+ "mask_token": "[MASK]",
10
+ "model_max_length": 512,
11
+ "never_split": null,
12
+ "pad_token": "[PAD]",
13
+ "sep_token": "[SEP]",
14
+ "strip_accents": null,
15
+ "tokenize_chinese_chars": true,
16
+ "tokenizer_class": "BertTokenizer",
17
+ "unk_token": "[UNK]"
18
+ }