Science Question Answering on Science QA (Task Acc. & Brier Delta)
65Task AccuracyFINCH
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| FINCHModel=Qwen3-4B2026.05 | 65 | — | |
| Opus-4.72026.07 | 63 | — | |
| DeepSeek-V4-Pro2026.07 | 63 | — | |
| SFTModel=Qwen3-4B2026.05 | 62.5 | — | |
| GPT-5.52026.07 | 60 | — | |
| FLOWModel=Qwen3-4B2026.05 | 58.3 | — | |
| Kimi-K2.62026.07 | 58 | — | |
| L2 RegModel=Qwen3-4B2026.05 | 57.5 | — | |
| WiSE-FTModel=Qwen3-4B2026.05 | 56.7 | — | |
| SciReasoner2026.07 | 55 | — | |
| LoRAModel=Qwen3-4B2026.05 | 51.7 | — | |
| STMModel=Qwen3-4B2026.05 | 49.2 | — | |
| TALRModel=Qwen3-4B2026.05 | 48.3 | — | |
| DFTModel=Qwen3-4B2026.05 | 46.7 | — | |
| BaseModel=Qwen3-4B2026.05 | 29.2 | — |