Open-ended Question Answering on OlymBench Phys v1 (test)
53.9Problem Level ScoreQwen3-VL-32B-Thinking
Evaluation results
| Method | Links | |
|---|---|---|
| Qwen3-VL-32B-Thinking2026.05 | 53.9 | |
| Claude Sonnet 4.5max_tokens=163842026.05 | 50.4 | |
| Physics-R1 (binary, 3-seed mean ±σ)max_tokens=16384, reward_type=binary, aggregation=3-seed mean2026.05 | 46.2 | |
| Physics-R1 (binary, seed 42)max_tokens=16384, reward_type=binary, seed=422026.05 | 45.4 | |
| Physics-R1 (dense)max_tokens=16384, reward_type=dense2026.05 | 40.5 | |
| Qwen3-VL-8B-Thinking (base)2026.05 | 39.3 | |
| Gemini 2.5 Pro2026.05 | 37.4 | |
| GPT-4o2026.05 | 19.7 | |
| InternVL3-8B2026.05 | 10.4 |