Scientific Reasoning on SciKnowEval Hardest
21Pass@1STV verifier
Evaluation Results
| Method | Links | |
|---|---|---|
| STV verifiergenerator=Qwen3-8B2026.05 | 21 | |
| Qwen3-32Bsize_multiplier=4x2026.05 | 8.9 | |
| Qwen3-235B-A22Bsize_multiplier=30x2026.05 | 8 | |
| No verifier traininggenerator=Qwen3-8B2026.05 | 2.1 | |
| No verificationgenerator=Qwen3-8B2026.05 | 1.5 |