Mathematical Reasoning on MATH (test) (Accuracy + Calibration Metrics)
15.02AccuracyEPICAR
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| EPICARBackbone=Llama-3-8B, Model Merging=true2026.01 | 15.02 | 0.571 | 0.443 | 0.361 | 0.328 | |
| EPICARBackbone=Llama-3-8B2026.01 | 14.42 | 0.595 | 0.415 | 0.362 | 0.298 | |
| EPICARBackbone=Llama-3-8B, Slow Thinking (ICL)=true2026.01 | 14.38 | 0.435 | 0.39 | 0.361 | 0.282 | |
| STaRBackbone=Llama-3-8B, Model Merging=true2026.01 | 13.72 | 0.555 | 0.492 | 0.381 | 0.365 | |
| STaRBackbone=Llama-3-8B2026.01 | 13.46 | 0.57 | 0.494 | 0.381 | 0.365 | |
| Base ModelBackbone=Llama-3-8B2026.01 | 13.3 | 0.544 | 0.496 | 0.384 | 0.368 | |
| STaRBackbone=Llama-3-8B, Slow Thinking (ICL)=true2026.01 | 13.06 | 0.415 | 0.436 | 0.375 | 0.312 | |
| Base ModelBackbone=Llama-3-8B, Slow Thinking (ICL)=true2026.01 | 12.22 | 0.387 | 0.448 | 0.384 | 0.316 | |
| EPICARBackbone=Llama-3-3B2026.01 | 8.58 | 0.568 | 0.108 | 0.053 | 0.097 | |
| EPICARBackbone=Llama-3-3B, Model Merging=true2026.01 | 7.86 | 0.593 | 0.167 | 0.05 | 0.106 | |
| STaRBackbone=Llama-3-3B, Model Merging=true2026.01 | 7.6 | 0.562 | 0.377 | 0.318 | 0.217 | |
| Base ModelBackbone=Llama-3-3B2026.01 | 7.56 | 0.555 | 0.376 | 0.315 | 0.216 | |
| STaRBackbone=Llama-3-3B2026.01 | 7.38 | 0.562 | 0.382 | 0.344 | 0.219 | |
| EPICARBackbone=Llama-3-3B, Slow Thinking (ICL)=true2026.01 | 6.46 | 0.571 | 0.44 | 0.436 | 0.257 | |
| STaRBackbone=Llama-3-3B, Slow Thinking (ICL)=true2026.01 | 6.24 | 0.548 | 0.609 | 0.456 | 0.432 | |
| Base ModelBackbone=Llama-3-3B, Slow Thinking (ICL)=true2026.01 | 5.84 | 0.529 | 0.605 | 0.46 | 0.424 | |
| STaRBackbone=Llama-3-1B, Model Merging=true2026.01 | 3.54 | 0.518 | 0.838 | 0.515 | 0.737 | |
| EPICARBackbone=Llama-3-1B, Model Merging=true2026.01 | 3.53 | 0.555 | 0.848 | 0.514 | 0.754 | |
| STaRBackbone=Llama-3-1B2026.01 | 3.46 | 0.491 | 0.838 | 0.515 | 0.737 | |
| Base ModelBackbone=Llama-3-1B2026.01 | 3.3 | 0.525 | 0.841 | 0.516 | 0.74 | |
| EPICARBackbone=Llama-3-1B2026.01 | 3.3 | 0.573 | 0.871 | 0.532 | 0.8 | |
| EPICARBackbone=Llama-3-1B, Slow Thinking (ICL)=true2026.01 | 3.14 | 0.576 | 0.885 | 0.534 | 0.817 | |
| STaRBackbone=Llama-3-1B, Slow Thinking (ICL)=true2026.01 | 3.12 | 0.469 | 0.826 | 0.516 | 0.716 | |
| Base ModelBackbone=Llama-3-1B, Slow Thinking (ICL)=true2026.01 | 3 | 0.507 | 0.827 | 0.517 | 0.716 |