Commit History
5 days ago
-
Retrained with fp32 master weights: 0.8908 mean nDCG@10 on the nine Korean tasks (was 0.8800)
Verified signature0ca2319
14 days ago
-
Pin the figure legend to the plot's top-right corner
Verified signature9d03c9c -
Update evaluation: PublicHealthQA/Belebele re-evaluated on corrected BM25 pools
Verified signature- PublicHealthQA and BelebeleRetrieval were scored against another query's BM25 candidates (the benchmark's mteb 1.x pools used per-process query ids); both are re-evaluated for every model, Belebele on its kor_Hang-kor_Hang subset. - Tables, key numbers and the figure updated; KURE-Reranker models drawn as stars. - Model weights are unchanged.dd38901
25 days ago
-
Update evaluation: official Qwen3-Reranker and KaLM-Reranker-V1-Large-R2 added, seq-cls variants and ettin removed
Verified signature- Tables now use the official Qwen/Qwen3-Reranker-{0.6B,4B,8B} instead of the tomaarsen seq-cls conversions, add KaLM-Embedding/KaLM-Reranker-V1-Large-R2, and drop cross-encoder/ettin-reranker-1b-v1 (not multilingual). - New figure; caption notes the non-uniform x-axis. - PPS note: same ~550 sampled pairs per task for every model; KaLM uses sdpa. - Model weights are unchanged.53e42a4
26 days ago
-
Move the figure legend slightly left of the x=1000 tick
Verified signature815f7d4 -
Align the figure legend with the x=1000 tick
Verified signature381a631 -
Use full KURE-Reranker names in the figure and fix a missing space in the PPS note
Verified signature2ce4a4d -
Update KURE-Reranker-base weights (bf16), model card and assets
Verified signature5cbe128