Multi-Hop Question Answering on HotpotQA (EM, LLM)
45.7EMDAC
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| DACBackbone=Qwen3-8B, FT=LoRA2026.06 | 45.7 | 62.8 | |
| Search-R1 (GRPO/OF)Backbone=Qwen3-8B, FT=Full2026.06 | 45 | 62.3 | |
| Search-R1 (GRPO/O)Backbone=Qwen3-8B, FT=Full2026.06 | 44.8 | 61 | |
| Search-R1 (PPO/OF)Backbone=Qwen3-8B, FT=Full2026.06 | 42.2 | 58.3 | |
| DACBackbone=Qwen2.5-7B-Instruct, FT=LoRA2026.06 | 40.3 | 57 | |
| Search-R1 (PPO/O)Backbone=Qwen3-8B, FT=Full2026.06 | 40.3 | 55.6 | |
| Search-R1 (GRPO/O)Backbone=Qwen2.5-7B-Instruct, FT=Full2026.06 | 38.5 | 54.3 | |
| Search-R1 (PPO/O)Backbone=Qwen2.5-7B-Instruct, FT=Full2026.06 | 38 | 54 | |
| Search-R1 (GRPO/OF)Backbone=Qwen2.5-7B-Instruct, FT=Full2026.06 | 37.1 | 53 | |
| IRCoTBackbone=Qwen3-8B, FT=-2026.06 | 36.9 | 53.5 | |
| Search-R1 (PPO/OF)Backbone=Qwen2.5-7B-Instruct, FT=Full2026.06 | 35.5 | 51.2 | |
| Search-o1Backbone=Qwen3-8B, FT=-2026.06 | 29.9 | 51.3 | |
| Naive RAGBackbone=Qwen2.5-7B-Instruct, FT=-2026.06 | 29.3 | 42.7 | |
| Search-o1Backbone=Qwen2.5-7B-Instruct, FT=-2026.06 | 29 | 44.8 | |
| Naive RAGBackbone=Qwen3-8B, FT=-2026.06 | 29 | 42.9 | |
| CoTBackbone=Qwen3-8B, FT=-2026.06 | 21.3 | 31.3 | |
| CoTBackbone=Qwen2.5-7B-Instruct, FT=-2026.06 | 20.8 | 30.5 | |
| DirectBackbone=Qwen2.5-7B-Instruct, FT=-2026.06 | 19.1 | 28 | |
| IRCoTBackbone=Qwen2.5-7B-Instruct, FT=-2026.06 | 18.7 | 40.9 | |
| DirectBackbone=Qwen3-8B, FT=-2026.06 | 18.2 | 28.1 |