jueri/longeval-2026-snapshot-3-index-dense

LongEval 2026 Snapshot 3 Terrier Dense Index

Description

This is a Terrier Dense index for the snapshot-3 of the LongEval 2026 test collection. It indexes the titles and abstracts of the scientific documents using Qwen/Qwen3-Embedding-4B and the default prompt.

# Load the artifact
import pyterrier as pt
import pyterrier_dr

index = pt.Artifact.from_hf('jueri/longeval-2026-snapshot-3-index-dense')
model = pyterrier_dr.SBertBiEncoder('Qwen/Qwen3-Embedding-4B')

retriever = model.query_encoder() >> index.retriever()

results = retriever.search("capital of Germany")

Benchmarks

TODO: Provide benchmarks for the artifact.

Reproduction

The index was created through the PyTerrier dense baseline.

Metadata

{
  "type": "sparse_index",
  "format": "terrier",
  "package_hint": "python-terrier"
}