List your company

Question Answering on MuSiQue (F1 score)

81.3F1 ScoreTri-RAG
8.29227.24646.265.154Aug 8, 2025Sep 25, 2025Nov 13, 2025Jan 1, 2026Feb 19, 2026Apr 9, 2026May 28, 2026

Evaluation results

MethodLinks
Tri-RAGGenerator Backbone=Qwen-Long2026.0481.3
AflowGenerator Backbone=Qwen-Long2026.0478.9
Tri-RAGGenerator Backbone=GPT-4o-mini2026.0474.6
Self-ActGenerator Backbone=Qwen-Long2026.0472.3
AflowGenerator Backbone=GPT-4o-mini2026.0471.6
Tri-RAGGenerator Backbone=GPT-4.1-nano2026.0471.2
AflowGenerator Backbone=GPT-4.1-nano2026.0469.4
Tri-RAGGenerator Backbone=Qwen-Max2026.0468.3
Self-ActGenerator Backbone=GPT-4o-mini2026.0467.3
Tri-RAGGenerator Backbone=Gemma-2-27B2026.0467.1
AflowGenerator Backbone=Qwen-Max2026.0465.3
AflowGenerator Backbone=Gemma-2-27B2026.0463.4
Self-ActGenerator Backbone=Qwen-Max2026.0461.8
Self-ActGenerator Backbone=GPT-4.1-nano2026.0461.1
Self-ActGenerator Backbone=Gemma-2-27B2026.0459.7
ReActGenerator Backbone=Qwen-Long2026.0458.9
DirectGenerator Backbone=Qwen-Long2026.0455.4
ReActGenerator Backbone=GPT-4o-mini2026.0453.6
ReActGenerator Backbone=Qwen-Max2026.0453.6
ReActGenerator Backbone=Gemma-2-27B2026.0451.7
DirectGenerator Backbone=Qwen-Max2026.0449.9
DirectGenerator Backbone=Gemma-2-27B2026.0448.4
ReActGenerator Backbone=GPT-4.1-nano2026.0447.3
GPT-4o (2024-11-20)Version=2024-11-202025.0846.04
DirectGenerator Backbone=GPT-4o-mini2026.0440.1
AGENT-RADAR2026.0539.72
DirectGenerator Backbone=GPT-4.1-nano2026.0438.2
AgentDropout2026.0537.06
AgentPrune2026.0537.05
CoT-SCsample_size (n)=82026.0536.33
Vanilla2026.0535.47
Compaction2026.0534.43
StepSearchBackbone=Qwen2.5-7B Instruct2026.0531.2
PiCABackbone=Qwen2.5-7B Instruct2026.0528.7
TIPSBackbone=Qwen2.5-7B Instruct2026.0527
StepSearchBackbone=Qwen2.5-3B Instruct2026.0526.1
Search-R1Backbone=Qwen2.5-7B Instruct2026.0525.6
PiCABackbone=Qwen2.5-3B Instruct2026.0523.5
Llama + LoRRBackbone=Llama3.2-3B, Optimization=LoRR2025.0821.59
Llama + DPOBackbone=Llama3.2-3B, Optimization=DPO2025.0820.89
ZeroSearchBackbone=Qwen2.5-7B Instruct2026.0520.4
Qwen + LoRRBackbone=Qwen2.5-3B, Optimization=LoRR2025.0818.4
Qwen (Base)Backbone=Qwen2.5-3B, Mode=Base2025.0818.28
Qwen + DPOBackbone=Qwen2.5-3B, Optimization=DPO2025.0818.03
Llama (Base)Backbone=Llama3.2-3B, Mode=Base2025.0817.21
Search-R1Backbone=Qwen2.5-3B Instruct2026.0517.1
TIPSBackbone=Qwen2.5-3B Instruct2026.0515.9
ZeroSearchBackbone=Qwen2.5-3B Instruct2026.0514.5
RAGBackbone=Qwen2.5-7B Instruct2026.0514.2
RAGBackbone=Qwen2.5-3B Instruct2026.0513.5
Search-o1Backbone=Qwen2.5-7B Instruct2026.0512.7
MT-PPOBackbone=Qwen2.5-7B Instruct2026.0512.1
Search-o1Backbone=Qwen2.5-3B Instruct2026.0511.7
MT-PPOBackbone=Qwen2.5-3B Instruct2026.0511.1