List your company

Multi-Hop Question Answering on MuSiQue (EM & LLM)

21.7EMSearch-R1 (GRPO/OF)
2.6687.60912.5517.491Jun 9, 2026

Evaluation results

MethodLinks
Search-R1 (GRPO/OF)Backbone=Qwen3-8B, FT=Full2026.0621.729.6
DACBackbone=Qwen3-8B, FT=LoRA2026.0621.128.5
Search-R1 (PPO/OF)Backbone=Qwen3-8B, FT=Full2026.0619.426.2
Search-R1 (GRPO/O)Backbone=Qwen3-8B, FT=Full2026.0618.326.1
DACBackbone=Qwen2.5-7B-Instruct, FT=LoRA2026.0617.223.7
Search-R1 (PPO/O)Backbone=Qwen3-8B, FT=Full2026.0616.522
Search-R1 (GRPO/OF)Backbone=Qwen2.5-7B-Instruct, FT=Full2026.0615.823.1
Search-R1 (PPO/O)Backbone=Qwen2.5-7B-Instruct, FT=Full2026.0615.621.6
Search-R1 (GRPO/O)Backbone=Qwen2.5-7B-Instruct, FT=Full2026.0615.621.8
Search-R1 (PPO/OF)Backbone=Qwen2.5-7B-Instruct, FT=Full2026.0612.719.6
IRCoTBackbone=Qwen3-8B, FT=-2026.061217.7
Search-o1Backbone=Qwen2.5-7B-Instruct, FT=-2026.0611.618.5
Search-o1Backbone=Qwen3-8B, FT=-2026.0611.322.1
Naive RAGBackbone=Qwen3-8B, FT=-2026.06711
Naive RAGBackbone=Qwen2.5-7B-Instruct, FT=-2026.066.49.9
CoTBackbone=Qwen3-8B, FT=-2026.065.810.9
CoTBackbone=Qwen2.5-7B-Instruct, FT=-2026.065.310.2
IRCoTBackbone=Qwen2.5-7B-Instruct, FT=-2026.064.811.1
DirectBackbone=Qwen2.5-7B-Instruct, FT=-2026.064.37.9
DirectBackbone=Qwen3-8B, FT=-2026.063.47.8