Synthetic Reasoning on Reasoning Primitives
55.46AccuracyMIDAS
Evaluation Results
| Method | Links | |
|---|---|---|
| MIDASModel Size=1.7B, Cooldown Protocol=Math, Five-shot prompting=true2025.12 | 55.46 | |
| LIDASModel Size=1.7B, Cooldown Protocol=Math, Five-shot prompting=true2025.12 | 53 | |
| LN-ScalingModel Size=1.7B, Cooldown Protocol=Math, Five-shot prompting=true2025.12 | 50.58 | |
| LIDASModel Size=360M, Cooldown Protocol=Math, Five-shot prompting=true2025.12 | 50.36 | |
| LIDASModel Size=1.7B, Cooldown Protocol=Standard, Five-shot prompting=true2025.12 | 47.34 | |
| LN-ScalingModel Size=1.7B, Cooldown Protocol=Standard, Five-shot prompting=true2025.12 | 44.38 | |
| BaselineModel Size=1.7B, Cooldown Protocol=Math, Five-shot prompting=true2025.12 | 42.77 | |
| LN-ScalingModel Size=360M, Cooldown Protocol=Math, Five-shot prompting=true2025.12 | 41.26 | |
| MIDASModel Size=1.7B, Cooldown Protocol=Standard, Five-shot prompting=true2025.12 | 40.88 | |
| MIDASModel Size=360M, Cooldown Protocol=Math, Five-shot prompting=true2025.12 | 35.14 | |
| BaselineModel Size=1.7B, Cooldown Protocol=Standard, Five-shot prompting=true2025.12 | 34.84 | |
| BaselineModel Size=360M, Cooldown Protocol=Math, Five-shot prompting=true2025.12 | 33.12 | |
| LN-ScalingModel Size=360M, Cooldown Protocol=Standard, Five-shot prompting=true2025.12 | 31.38 | |
| LIDASModel Size=360M, Cooldown Protocol=Standard, Five-shot prompting=true2025.12 | 31.2 | |
| BaselineModel Size=360M, Cooldown Protocol=Standard, Five-shot prompting=true2025.12 | 30.06 | |
| MIDASModel Size=360M, Cooldown Protocol=Standard, Five-shot prompting=true2025.12 | 28.18 |