List your company

Multi-Hop Question Answering on HotpotQA (token-level F1)

62.6Token-level F1Oracle
17.98429.56741.1552.733Apr 19, 2026

Evaluation results

MethodLinks
OracleMain agent backbone=Qwen2.5-7B-Instruct2026.0462.6
COSEARCHMain agent backbone=Qwen2.5-7B-Instruct2026.0459.5
OracleMain agent backbone=Qwen2.5-3B-Instruct2026.0458.9
Fixed RankerMain agent backbone=Qwen2.5-7B-Instruct2026.0457.8
Search-R1Main agent backbone=Qwen2.5-7B-Instruct2026.0457.6
ZeroSearchMain agent backbone=Qwen2.5-7B-Instruct2026.0457.5
Retrieval OnlyMain agent backbone=Qwen2.5-7B-Instruct2026.0457.2
COSEARCHMain agent backbone=Qwen2.5-3B-Instruct2026.0447.8
Fixed RankerMain agent backbone=Qwen2.5-3B-Instruct2026.0447.4
ZeroSearchMain agent backbone=Qwen2.5-3B-Instruct2026.0445.5
Retrieval OnlyMain agent backbone=Qwen2.5-3B-Instruct2026.0445.4
Search-R1Main agent backbone=Qwen2.5-3B-Instruct2026.0445
RAGMain agent backbone=Qwen2.5-7B-Instruct2026.0437.1
RAGMain agent backbone=Qwen2.5-3B-Instruct2026.0437.1
Search-o1Main agent backbone=Qwen2.5-7B-Instruct2026.0431.6
Direct InferenceMain agent backbone=Qwen2.5-7B-Instruct2026.0429.2
CoTMain agent backbone=Qwen2.5-7B-Instruct2026.0424.4
Search-o1Main agent backbone=Qwen2.5-3B-Instruct2026.0422.5
Direct InferenceMain agent backbone=Qwen2.5-3B-Instruct2026.0420.2
CoTMain agent backbone=Qwen2.5-3B-Instruct2026.0419.7