Scientific Reasoning on UGPhysics English
33.1Overall AccuracySci-CoE-Stage 2
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Sci-CoE-Stage 2Backbone=Qwen3-8B, Data Scale=30k2026.02 | 33.1 | 30.51 | 34.8 | 34.99 | |
| Sci-CoE-Stage 2Backbone=Qwen3-8B, Data Scale=18k2026.02 | 32.46 | 30.21 | 33.3 | 34.38 | |
| Sci-CoE-Stage 1Backbone=Qwen3-8B, Data Scale=4k2026.02 | 32.03 | 30.25 | 30.62 | 34.38 | |
| Qwen3-8B Base ModelData Scale=None2026.02 | 31.76 | 30.73 | 29.98 | 33.51 | |
| Mistral-Small-Instruct-2409Data Scale=None2026.02 | 25.72 | 22.71 | 22.7 | 29.97 | |
| Sci-CoE-Stage 2Backbone=Qwen2.5-7B-Instruct, Data Scale=30k2026.02 | 22.64 | 21.84 | 23.13 | 24.91 | |
| Sci-CoE-Stage 2Backbone=Qwen2.5-7B-Instruct, Data Scale=18k2026.02 | 21.92 | 20.92 | 21.31 | 23.17 | |
| Sci-CoE-Stage 1Backbone=Qwen2.5-7B-Instruct, Data Scale=4k2026.02 | 21.07 | 20.14 | 19.81 | 22.51 | |
| Qwen2.5-7B-Instruct Base ModelData Scale=None2026.02 | 20.67 | 18.88 | 18.52 | 23.34 | |
| Yi-1.5-9B-ChatData Scale=None2026.02 | 17.61 | 16 | 15.85 | 19.94 | |
| Mathstral-7B-v0.1Data Scale=None2026.02 | 17.45 | 14.82 | 17.77 | 19.94 | |
| Ministral-8B-Instruct-2410Data Scale=None2026.02 | 16.39 | 13.95 | 15.52 | 19.2 | |
| Llama-3.1-8B-InstructData Scale=None2026.02 | 14.66 | 12.64 | 14.35 | 16.8 |