Premise Selection on Isabelle (IS) (test)
86.3Exact Match AccuracyLlama-3.1 (RULES)
Evaluation Results
| Method | Links | |
|---|---|---|
| Llama-3.1 (RULES)Optimization=RULES2026.02 | 86.3 | |
| Llama-3.1 (GRPO)Optimization=GRPO2026.02 | 85.9 | |
| Qwen2.5 (RULES)Optimization=RULES2026.02 | 81.3 | |
| OLMO2 (RULES)Optimization=RULES2026.02 | 76.6 | |
| DeepMath (RULES)Optimization=RULES2026.02 | 75.8 | |
| Qwen2.5 (GRPO)Optimization=GRPO2026.02 | 74.2 | |
| OLMO2 (GRPO)Optimization=GRPO2026.02 | 72.2 | |
| DeepMath (GRPO)Optimization=GRPO2026.02 | 67.6 | |
| Qwen2.5 (vanilla)Optimization=vanilla2026.02 | 55.5 | |
| Llama-3.1 (vanilla)Optimization=vanilla2026.02 | 54.7 | |
| OLMO2 (vanilla)Optimization=vanilla2026.02 | 53.9 | |
| DeepMath (vanilla)Optimization=vanilla2026.02 | 51.6 |