First Integral Discovery on Hard (test)
63.7AccuracyFISolver-Base + RL(HS)(LD)
Evaluation results
| Method | Links | |
|---|---|---|
| FISolver-Base + RL(HS)(LD)Parameter Scale=1.5B, Beam Size=70, Tokens Per Sample=1611.8 (23.0 × 70)2026.05 | 63.7 | |
| DeepSeek-V3.2-ExpParameter Scale=684.7B, Beam Size=1, Tokens Per Sample=2869.62026.05 | 42.8 |