List your company

Multi-Hop Question Answering on 2Wiki (token-level F1)

58.7Token-Level F1COSEARCH
10.96423.35735.7548.143Apr 19, 2026

Evaluation results

MethodLinks
COSEARCHMain agent backbone=Qwen2.5-7B-Instruct2026.0458.7
OracleMain agent backbone=Qwen2.5-7B-Instruct2026.0458.4
Fixed RankerMain agent backbone=Qwen2.5-7B-Instruct2026.0456
ZeroSearchMain agent backbone=Qwen2.5-7B-Instruct2026.0455.9
Retrieval OnlyMain agent backbone=Qwen2.5-7B-Instruct2026.0455.6
Search-R1Main agent backbone=Qwen2.5-7B-Instruct2026.0454.7
OracleMain agent backbone=Qwen2.5-3B-Instruct2026.0451
COSEARCHMain agent backbone=Qwen2.5-3B-Instruct2026.0442.8
Fixed RankerMain agent backbone=Qwen2.5-3B-Instruct2026.0441.1
Retrieval OnlyMain agent backbone=Qwen2.5-3B-Instruct2026.0440.6
Search-R1Main agent backbone=Qwen2.5-3B-Instruct2026.0439.7
ZeroSearchMain agent backbone=Qwen2.5-3B-Instruct2026.0438.7
Direct InferenceMain agent backbone=Qwen2.5-7B-Instruct2026.0430.4
Search-o1Main agent backbone=Qwen2.5-7B-Instruct2026.0428.6
CoTMain agent backbone=Qwen2.5-7B-Instruct2026.0426.4
Direct InferenceMain agent backbone=Qwen2.5-3B-Instruct2026.0424.9
RAGMain agent backbone=Qwen2.5-7B-Instruct2026.0424.4
CoTMain agent backbone=Qwen2.5-3B-Instruct2026.0422.9
RAGMain agent backbone=Qwen2.5-3B-Instruct2026.0419.3
Search-o1Main agent backbone=Qwen2.5-3B-Instruct2026.0412.8