Question Answering on HotpotQA (val)
57.3F1 ScoreM-ASK iterative search
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| M-ASK iterative searchModel Size=7B2026.05 | 57.3 | — | — | — | — | |
| QD-Retrieve-AnswerModel Size=14B2026.05 | 54.2 | — | — | — | — | |
| M-ASK iterative searchModel Size=3B2026.05 | 52.9 | — | — | — | — | |
| M-ASK iterative searchModel Size=1.5B2026.05 | 52.5 | — | — | — | — | |
| QD-Retrieve-AnswerModel Size=7B2026.05 | 52.3 | — | — | — | — | |
| QD-Retrieve-Evidence-AnswerModel Size=14B2026.05 | 52.2 | — | — | — | — | |
| QD-Retrieve-AnswerModel Size=1.5B2026.05 | 49.3 | — | — | — | — | |
| QD-Retrieve-AnswerModel Size=3B2026.05 | 49.3 | — | — | — | — | |
| QD-Retrieve-Evidence-AnswerModel Size=7B2026.05 | 48.6 | — | — | — | — | |
| QD-Retrieve-Evidence-AnswerModel Size=3B2026.05 | 47.3 | — | — | — | — | |
| QD-Retrieve-Evidence-AnswerModel Size=1.5B2026.05 | 43.5 | — | — | — | — | |
| IRCoTpipeline=FlashRAG2025.05 | 41.5 | — | — | — | — | |
| QD-Retrieve-AnswerModel Size=0.5B2026.05 | 39.7 | — | — | — | — | |
| QD-Retrieve-Evidence-AnswerModel Size=0.5B2026.05 | 35.9 | — | — | — | — | |
| Standard RAGpipeline=FlashRAG2025.05 | 35.3 | — | — | — | — | |
| SuRepipeline=FlashRAG2025.05 | 33.4 | — | — | — | — | |
| REPLUGpipeline=FlashRAG2025.05 | 31.2 | — | — | — | — | |
| FLAREpipeline=FlashRAG2025.05 | 28 | — | — | — | — | |
| M-ASK iterative searchModel Size=0.5B2026.05 | 22.6 | — | — | — | — | |
| Llama-3.1-8B-InstN=7405, Decoding=stochastic2026.02 | — | 70.4 | 0.675 | -0.277 | 0.675 | |
| Llama-3.2-1B-InstN=7405, Decoding=stochastic2026.02 | — | 48.3 | 0.594 | -0.163 | 0.594 | |
| Llama-3.2-3B-InstN=7405, Decoding=stochastic2026.02 | — | 63.2 | 0.662 | -0.271 | 0.662 | |
| Qwen2.5-0.5B-InstN=7405, Decoding=stochastic2026.02 | — | 41 | 0.568 | -0.116 | 0.575 | |
| Qwen2.5-1.5B-InstN=7405, Decoding=stochastic2026.02 | — | 53.7 | 0.603 | -0.179 | 0.603 | |
| Qwen2.5-7B-InstN=7405, Decoding=stochastic2026.02 | — | 67.5 | 0.654 | -0.25 | 0.654 |