Compositional Reasoning on ARC-AGI 2
33.6AccuracyRCE
Evaluation Results
| Method | Links | |
|---|---|---|
| RCEModel=Qwen-14B2026.02 | 33.6 | |
| RCEModel=Llama-3-8B2026.02 | 29.8 | |
| RCEModel=Mistral-7B2026.02 | 28 | |
| DisCOModel=Mistral-7B2026.02 | 19.7 | |
| BaseModel=Qwen-14B2026.02 | 19.3 | |
| GRPOModel=Mistral-7B2026.02 | 18.2 | |
| ToTModel=Mistral-7B2026.02 | 17.3 | |
| SCModel=Mistral-7B, Samples (n)=162026.02 | 16.8 | |
| CoTModel=Mistral-7B2026.02 | 15.1 | |
| BaseModel=Llama-3-8B2026.02 | 14.1 | |
| BaseModel=Mistral-7B2026.02 | 12.4 |