List your company

Question Answering on Musique

26EMICRL
-0.8326.13413.120.066Jan 21, 2026Feb 4, 2026Feb 19, 2026Mar 6, 2026Mar 21, 2026Apr 5, 2026Apr 20, 2026

Evaluation results

MethodLinks
ICRLModel=Qwen2.5-7B2026.0326—
SLATEBackbone=Qwen2.5-7B-Base2026.0224.7—
StepSearchBackbone=Qwen2.5-7B-Base, Training Data=MuSiQue (19k)2026.0221.6—
SLATEBackbone=Qwen2.5-3B-Base, Training Data=NQ+HotpotQA2026.0221.6—
ICRLModel=Qwen2.5-3B2026.0320—
ParallelSearchModel=Qwen2.5-7B2026.0319.7—
SEARCH-R1Backbone=Qwen2.5-7B-Base2026.0219.6—
LAnR-BaseRetrieval Strategy=w/ Multi-Hop Retrieval, Model Variant=Base2026.0418.7—
ZeroSearchModel=Qwen2.5-7B2026.0318.4—
StepSearchBackbone=Qwen2.5-3B-Base, Training Data=MuSiQue (19k), Knowledge Base=Wiki-182026.0218.1—
AutoRefine-InstructRetrieval Strategy=w/ Multi-Hop Retrieval, Model Variant=Instruct2026.0416.9—
AutoRefine-BaseRetrieval Strategy=w/ Multi-Hop Retrieval, Model Variant=Base2026.0415.7—
Search-R1Model=Qwen2.5-7B2026.0314.6—
ReSearchBackbone=Qwen2.5-7B-Base2026.0214.3—
Search-R1-InstructRetrieval Strategy=w/ Multi-Hop Retrieval, Model Variant=Instruct2026.0412.4—
Reject SamplingModel=Qwen2.5-7B2026.0312.3—
Search-R1Model=Qwen2.5-3B2026.0310.3—
ZeroSearchBackbone=Qwen2.5-7B-Base2026.0210.2—
ZeroSearchModel=Qwen2.5-3B2026.039.8—
ReSearch-InstructRetrieval Strategy=w/ Multi-Hop Retrieval, Model Variant=Instruct2026.049.5—
R1-baseModel=Qwen2.5-7B2026.038.3—
AdaTIRBackbone=Qwen2.5-3b-Base2026.017.699
R1 (no search)Backbone=Qwen2.5-7B-Base2026.027.6—
ReSearchBackbone=Qwen2.5-3B-Base2026.027.4—
ReSearch-BaseRetrieval Strategy=w/ Multi-Hop Retrieval, Model Variant=Base2026.047.4—
IRCoTModel=Qwen2.5-7B2026.037.2—
R1-instructModel=Qwen2.5-7B2026.037.2—
IRCoTBackbone=Qwen2.5-7B-Base2026.027.2—
IRCoTModel=Qwen2.5-3B2026.036.7—
IRCoTRetrieval Strategy=w/ Multi-Hop Retrieval2026.046.7—
SFTModel=Qwen2.5-7B2026.036.6—
Search-R1-BaseRetrieval Strategy=w/ Multi-Hop Retrieval, Model Variant=Base2026.046.6—
RAGBackbone=Qwen2.5-7B-Base2026.026.1—
R1-instructModel=Qwen2.5-3B2026.036—
Reject SamplingModel=Qwen2.5-3B2026.035.9—
IRCoTBackbone=Qwen2.5-3B-Base2026.025.9—
Search-o1Model=Qwen2.5-7B2026.035.8—
RAGModel=Qwen2.5-7B2026.035.8—
ZeroSearchBackbone=Qwen2.5-3B-Base2026.025.6—
Search-R1Backbone=Qwen2.5-3b-Base2026.015.4100
Search-o1Model=Qwen2.5-3B2026.035.4—
SFTBackbone=Qwen2.5-7B-Base2026.025.4—
Search-o1Retrieval Strategy=w/ Multi-Hop Retrieval2026.045.4—
Search-o1Backbone=Qwen2.5-7B-Base2026.025.3—
RAGBackbone=Qwen2.5-3B-Base2026.025.1—
R1 (no search)Backbone=Qwen2.5-3B-Base2026.025.1—
SEARCH-R1Backbone=Qwen2.5-3B-Base, Training Data=NQ+HotpotQA2026.024.9—
RAGModel=Qwen2.5-3B2026.034.7—
Naive RAGRetrieval Strategy=w/ Single-Hop Retrieval2026.044.7—
CoTBackbone=Qwen2.5-7B-Base2026.024.5—
Search-o1Backbone=Qwen2.5-3B-Base2026.024.5—
SFTModel=Qwen2.5-3B2026.034.4—
SFTRetrieval Strategy=w/o Retrieval2026.044.4—
SFTBackbone=Qwen2.5-3B-Base2026.023.9—
CoTBackbone=Qwen2.5-3B-Base2026.023.2—
DirectModel=Qwen2.5-7B2026.033.1—
DirectBackbone=Qwen2.5-7B-Base2026.022.7—
CoTModel=Qwen2.5-7B2026.032.2—
DirectModel=Qwen2.5-3B2026.032—
Direct GenerationRetrieval Strategy=w/o Retrieval2026.042—
DirectBackbone=Qwen2.5-3B-Base2026.021.8—
CoTModel=Qwen2.5-3B2026.030.2—