Reasoning on ARC-AGI 2
50AccuracyAdaEvolve
Evaluation Results
| Method | Links | |
|---|---|---|
| AdaEvolveBackbone=Gemini-3-Pro, Inference budget=matched2026.02 | 50 | |
| AdaEvolveBackbone=GPT-5, Inference budget=matched2026.02 | 49 | |
| OpenEvolveBackbone=Gemini-3-Pro, Inference budget=matched2026.02 | 44 | |
| OpenEvolveBackbone=GPT-5, Inference budget=matched2026.02 | 42 | |
| Claude Sonnet 4.52025.11 | 5.3 | |
| Grok 42025.11 | 2.7 | |
| Gemini 2.5 Pro2025.11 | 1.9 | |
| Sora-22025.11 | 1.3 | |
| GPT-5 high2025.11 | 0.5 |