Multi-hop Question Answering on HotpotQA distractor setting
52.01ConcisenessQwen3
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Qwen3Size=14B2026.01 | 52.01 | 73.18 | 97.31 | |
| Qwen3Size=30B-A3B2026.01 | 50.97 | 77.09 | 97.77 | |
| Qwen3Size=4B2026.01 | 50.39 | 75.58 | 96.87 | |
| Qwen3Size=1.7B2026.01 | 50.07 | 66.17 | 92.9 | |
| DeepSeek-R1Size=671B2026.01 | 49.6 | 73.91 | 98.56 | |
| GPT-5.1Size=UNK2026.01 | 48.37 | 72.61 | 99.15 | |
| Qwen3Size=8B2026.01 | 47.47 | 75.36 | 97.68 | |
| GPT-o3Size=UNK2026.01 | 47.44 | 73.24 | 98.92 | |
| Qwen3Size=235B-A22B2026.01 | 45.62 | 73.8 | 98.87 | |
| GPT-4oSize=UNK2026.01 | 44.9 | 75.57 | 97.36 | |
| DeepSeek-V3.2Size=671B2026.01 | 44.81 | 76.34 | 97.96 | |
| Gemini-3-proSize=UNK2026.01 | 39.69 | 78.65 | 98.56 | |
| LLaMA-3.1Size=70B2026.01 | 39.4 | 71.4 | 96.04 | |
| SelfCiteSize=8B2026.01 | 39.23 | 66.74 | 91.04 | |
| LongCiteSize=8B2026.01 | 38.85 | 64.56 | 91.75 | |
| Claude-4.5Size=UNK2026.01 | 36.4 | 78.33 | 98.07 | |
| Qwen3Size=0.6B2026.01 | 33.11 | 52.16 | 81.52 | |
| FRONTSize=13B2026.01 | 27.76 | 57.96 | 82.86 | |
| LLaMA-3.1Size=8B2026.01 | 26.46 | 54.94 | 84.28 | |
| LLaMA-3.2Size=1B2026.01 | 10.39 | 17.45 | 46.45 | |
| LLaMA-3.2Size=3B2026.01 | 10.39 | 19.02 | 70.41 |