Question Answering on HotpotQA (test) (Accuracy, EM, and F1 Metrics)
37.623EMMHGPO-FoF(os)
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| MHGPO-FoF(os)Framework=LLM + MAS + MARL2025.06 | 37.623 | 41.512 | 50.858 | |
| MHGPO-RRFramework=LLM + MAS + MARL2025.06 | 37.043 | 40.864 | 49.724 | |
| R2-Searcher2026.06 | 37 | — | 47.6 | |
| MHGPO-FoFFramework=LLM + MAS + MARL2025.06 | 36.57 | 40.459 | 49.429 | |
| R1-Searcher(PPO)Framework=LLM + RAG (+RL)2025.06 | 35.982 | 37.493 | 47.232 | |
| MAPPOFramework=LLM + MAS + MARL2025.06 | 34.45 | 38.217 | 46.4 | |
| MHGPO-ISFramework=LLM + MAS + MARL2025.06 | 33.707 | 37.677 | 45.582 | |
| AutoRefine2026.06 | 33.6 | — | 43.8 | |
| R1-Searcher(GRPO)Framework=LLM + RAG (+RL)2025.06 | 32.673 | 36.92 | 45.392 | |
| ReSearch2026.06 | 30.8 | — | 41.1 | |
| Search-R12026.06 | 30.1 | — | 39 | |
| Qwen2.5-72BFramework=LLM2025.06 | 25.564 | 30.304 | 36.183 | |
| ZeroSearch2026.06 | 24.4 | — | 33.2 | |
| Qwen2.5-72BFramework=LLM + MAS2025.06 | 24.254 | 32.519 | 34.631 | |
| Search-o1Framework=LLM + RAG (+RL)2025.06 | 17.92 | 23.023 | 25.702 | |
| Search-o12026.06 | 15.5 | — | 22.5 | |
| RAG2026.06 | 14.2 | — | 19.3 | |
| Llama3.1-8BFramework=LLM2025.06 | 14.018 | 21.269 | 22.782 | |
| Llama3.1-8BFramework=LLM + MAS2025.06 | 14 | 20.8 | 21.452 | |
| Direct Inference2026.06 | 13.7 | — | 20.2 | |
| COT2026.06 | 13.6 | — | 19.7 | |
| IRCOT2026.06 | 13.2 | — | 19.9 | |
| PPOFramework=LLM + RL2025.06 | 1.201 | 25.434 | 24.521 | |
| GRPOFramework=LLM + RL2025.06 | 0.5 | 27.643 | 27.421 |