Dataset Viewer
Duplicate
The dataset viewer is not available for this split.
Cannot extract the features (columns) for the split 'train' of the config 'default' of the dataset.
Error code:   FeaturesError
Exception:    ArrowInvalid
Message:      JSON parse error: Invalid value. in row 0
Traceback:    Traceback (most recent call last):
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 342, in _generate_tables
                  df = pandas_read_json(f)
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 38, in pandas_read_json
                  return pd.read_json(path_or_buf, **kwargs)
                         ~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/pandas/io/json/_json.py", line 791, in read_json
                  json_reader = JsonReader(
                      path_or_buf,
                  ...<16 lines>...
                      engine=engine,
                  )
                File "/usr/local/lib/python3.14/site-packages/pandas/io/json/_json.py", line 905, in __init__
                  self.data = self._preprocess_data(data)
                              ~~~~~~~~~~~~~~~~~~~~~^^^^^^
                File "/usr/local/lib/python3.14/site-packages/pandas/io/json/_json.py", line 917, in _preprocess_data
                  data = data.read()
                File "/usr/local/lib/python3.14/site-packages/datasets/utils/file_utils.py", line 838, in read_with_retries
                  out = read(*args, **kwargs)
                File "<frozen codecs>", line 325, in decode
              UnicodeDecodeError: 'utf-8' codec can't decode byte 0xb5 in position 11: invalid start byte
              
              During handling of the above exception, another exception occurred:
              
              Traceback (most recent call last):
                File "/src/services/worker/src/worker/job_runners/split/first_rows.py", line 244, in compute_first_rows_from_streaming_response
                  iterable_dataset = iterable_dataset._resolve_features()
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 4645, in _resolve_features
                  features = _infer_features_from_batch(self.with_format(None)._head())
                                                        ~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2866, in _head
                  return next(iter(self.iter(batch_size=n)))
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 3073, in iter
                  for key, pa_table in ex_iterable.iter_arrow():
                                       ~~~~~~~~~~~~~~~~~~~~~~^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2581, in _iter_arrow
                  yield from self.ex_iterable._iter_arrow()
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 558, in _iter_arrow
                  for key, pa_table in iterator:
                                       ^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 440, in _iter_arrow
                  for key, pa_table in self.generate_tables_fn(**gen_kwags):
                                       ~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 345, in _generate_tables
                  raise e
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 291, in _generate_tables
                  pa_table = paj.read_json(
                      io.BytesIO(batch), read_options=paj.ReadOptions(block_size=block_size)
                  )
                File "pyarrow/_json.pyx", line 364, in pyarrow._json.read_json
                File "pyarrow/error.pxi", line 155, in pyarrow.lib.pyarrow_internal_check_status
                  return check_status(status)
                File "pyarrow/error.pxi", line 92, in pyarrow.lib.check_status
                  raise convert_status(status)
              pyarrow.lib.ArrowInvalid: JSON parse error: Invalid value. in row 0

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

PAMI: From Dimensions to Directions

Public figure data and source archives for From Dimensions to Directions: Discovering Latent Concepts in Large Language Models. Published at the author's request on 2026-11-16. The packaged research snapshot is dated 2026-10-06.

Files and layout

  • Full_Plotting_Workspace_Code_and_Data.zip: comprehensive local paper_space source and data snapshot (all code/text and JSON/CSV/NPZ/NPY/etc. inputs, including stats, subspace, task3 results and historical arrays). Extract into paper_space/. Reference PDFs, caches and generated figure exports are excluded. Its complete file list and hashes are recorded in GitHub PAPER_SPACE_MANIFEST.json.
  • PCA_Source_Data.zip: raw arrays, PCA parameters, plotting tables, historical archives and candidate upstream arrays. Extract to PCA_Source_Data/.
  • Figure_Source_and_Code.zip: complete existing delivery package, including current Fig.5/7 plotting data and scripts. Extract at your working root; keep its reproduction/ directory intact.
  • Original_Plotting_Supplement.zip: original plotting directory, including recursive original_code/figs_unified/ sources and local data. Extract separately into historical_reproduction/ to avoid replacing current code with older versions.
  • Figure2_Native_Sources.zip, Editable_Figures_and_Preview.zip: native editable artwork and figure outputs.
  • Wiki_Paper_Snapshot.zip: local wiki paper snapshot, excluding reference-paper caches and reviews.
  • Overleaf_Updated_Source.zip: historical Overleaf snapshot from the 2026-10-06 delivery. Not verified as the current remote project on 2026-11-16: the existing session returned HTTP 401.
  • Reproduction_README.txt: original current-entry-point instructions.
  • Recovery_Report_zh.pdf, Figure2_Source_Report_zh.pdf: provenance and source limitations.
  • SHA256SUMS.json: SHA-256 and exact byte sizes for every artifact.

Reproduction

Extract Figure_Source_and_Code.zip and install reproduction/requirements.txt, then run:

python reproduction/build_plots.py --smoke
python reproduction/build_plots.py
python verify_data.py

These are the archived instructions; no new full plotting run was performed during this upload. The source layout is retained. Data paths inside the current package resolve relative to reproduction/data/. Historical scripts may retain their original absolute input paths: consult their source map before running them.

Provenance limits

Fig.5 contains a measured 5-model by 5-token grid and saved PCA parameters. Fig.7a has matched native PCA data. Fig.7b preserves exact vector/display geometry; the high-dimensional arrays labelled candidates have not been established as pointwise matching sources for its displayed cloud. Historical author-set values and later recomputed results are retained with their original provenance; do not mix versions silently. Model checkpoints and external full training corpora are not included.

Source code and project

Browsable plotting code and path records are in the author's private GitHub repository: https://github.com/chenhangcuisg-code/overleaf-agent/tree/master/projects/channel-analysis-tpami The source ZIP here permits downloading the existing source delivery without access to that private repository. Overleaf project: https://www.overleaf.com/project/6a217afa1ba25b1965861fb9 Public availability does not assign a new license to third-party components; retain their original notices.

Original-result recovery (2026-11-17)

  • Missing_Source_Recovery_20261117.zip: 134 unchanged original/candidate files, including six-model K=1 real/null JSONs from Leonardo, 77 raw Qwen VL channel/token JSONs and raw lm-eval baseline/run outputs. Includes original high-dimensional candidates and damaged-file evidence, explicitly classified in the report.
  • Recovery_20261117_README.md: item-by-item FOUND / NOT FOUND, scope, exact paths/sizes and limitations.
  • Recovery_20261117_SOURCE_MANIFEST.json: original source paths, exact sizes, SHA-256, keys/first rows and archive layout.

The paper greedy Llama3-8B GSM8K baseline reports strict=71.3419%, flexible=77.8620%; a separate earlier baseline has flexible=78.3169%. Do not mix runs or substitute the strict score for flexible extraction. Gemma3-4B reports strict=75.6634%, flexible=75.8908%. No values were generated, recomputed or edited in this recovery. Exact Fig.7(b) coordinates, the missing per-token Llama GSM8K run and independent InternVL diversity inputs remain unresolved. Current Overleaf was not modified.

Downloads last month
64