Multi-Round Chat Retrieval on MRCR
92.3AccuracyQwenLong-L1.5-30B
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| QwenLong-L1.5-30BParameters=30B, Context Limit=256K2026.05 | 92.3 | — | — | — | — | — | — | |
| ACCBase Model=Qwen3-30B-A3B-Thinking2026.05 | 68.28 | — | — | — | — | — | — | |
| Qwen3-30B-A3B-ThinkingModel Role=Base Model, Parameters=30B2026.05 | 50.19 | — | — | — | — | — | — | |
| Qwen2.5-7B-LongPO-128KParameters=7B, Context Limit=128K2026.05 | 31.5 | — | — | — | — | — | — | |
| Qwen2.5-14B-LongRLVRParameters=14B2026.05 | 20.06 | — | — | — | — | — | — | |
| Qwen2.5-7B-LongRLVRParameters=7B2026.05 | 19.76 | — | — | — | — | — | — | |
| Falcon-H1R# Param.=7B2026.02 | — | 13.18 | 9.06 | 6.93 | 9.17 | 8.22 | 7.54 | |
| MiniCPM-SALA# Param.=9B2026.02 | — | 29.77 | 20.57 | 16.56 | 28.62 | 19.62 | 10.12 | |
| Ministral-3-R# Param.=8B2026.02 | — | 44.02 | 35.8 | 17.23 | 50.3 | 22.66 | 14.47 | |
| Nemotron-Nano-v2# Param.=9B2026.02 | — | 20.91 | 13.69 | 13.24 | 14.61 | 12.2 | 7.55 | |
| Qwen3# Param.=8B2026.02 | — | 29.2 | 21.56 | 17.82 | 26.5 | 14.75 | 12.15 |