Citation and Evidence Recall on MuSiQue
80.2RkGPT-5.2
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| GPT-5.2Reasoning condition=Reasoning2025.10 | 80.2 | 68.6 | 87 | |
| GPT-5-MiniReasoning condition=Reasoning2025.10 | 73.2 | 64.5 | 80.5 | |
| Oracle-DRAGReasoning condition=No Reasoning2025.10 | 70.4 | 100 | 70.4 | |
| GPT-OSS-120BReasoning condition=Reasoning2025.10 | 66.2 | 63.8 | 74.2 | |
| Phi-4Reasoning condition=No Reasoning2025.10 | 64.7 | 36.3 | 75.4 | |
| Qwen3-32BReasoning condition=Reasoning2025.10 | 62.5 | 54.4 | 76.1 | |
| Mistr-S-3.2-24BReasoning condition=No Reasoning2025.10 | 61.5 | 50 | 70.1 | |
| Qwen3-14BReasoning condition=Reasoning2025.10 | 57.8 | 51.3 | 69.7 | |
| Qwen3-8BReasoning condition=Reasoning2025.10 | 54.9 | 50.4 | 68 | |
| GPT-OSS-20BReasoning condition=Reasoning2025.10 | 51.1 | 48.3 | 63.3 | |
| Oracle-BM25Reasoning condition=No Reasoning2025.10 | 48.7 | 100 | 48.7 | |
| Llama-3.1-8BReasoning condition=No Reasoning2025.10 | 44.7 | 36.4 | 54.3 | |
| Qwen3-4BReasoning condition=Reasoning2025.10 | 44 | 34.7 | 63.2 | |
| Phi-4-MiniReasoning condition=No Reasoning2025.10 | 41.2 | 21.8 | 53.5 | |
| Llama-3.3-70BReasoning condition=No Reasoning2025.10 | 41.1 | 29.6 | 49.1 | |
| Llama-3.2-3BReasoning condition=No Reasoning2025.10 | 39.1 | 24.7 | 46 | |
| Qwen3-1.7BReasoning condition=Reasoning2025.10 | 31.8 | 29.7 | 40.9 | |
| Ministr-8BReasoning condition=No Reasoning2025.10 | 31.3 | 26.7 | 44 | |
| Qwen3-0.6BReasoning condition=Reasoning2025.10 | 30.5 | 1.1 | 50 | |
| Llama-3.2-1BReasoning condition=No Reasoning2025.10 | 10 | 11.5 | 14.1 |