RadLIT-9

Radiology education retrieval benchmark with 420 multiple-choice questions from 36 open-access review articles across nine ABR subspecialties.

5rows
mrrprimary metric
2026-05-06sampled

Metadata

ID
radlit9
Category
Retrieval
Release
Unknown

Metrics

MRR, nDCG@10, Recall@10

Latest Results

Rows are parsed from the RadLIT-9 Hugging Face dataset card leaderboard. Source display names are preserved.

Rank Subject MRR Model Match Provenance Sampled
1 RadLITE v2.0 0.83 — Imported 2026-05-06
2 BM25 (baseline) 0.82 — Imported 2026-05-06
3 RadLITx v1.5 0.78 — Imported 2026-05-06
4 RadLIT bi-encoder 0.69 — Imported 2026-05-06
5 BGE-large-en-v1.5 0.67 — Imported 2026-05-06