20.7% (207/1000) on ARC-AGI-2 Training SetZero neural networks. Zero LLMs. Every solution is a verifiable symbolic program.
GitHub | Website | X
Run the full evaluation on all 1000 ARC-AGI-2 training tasks.
Verantyx V6 by kofdai | Pure symbolic program synthesis โ no neural networks, no LLMsGitHub | HuggingFace | verantyx.ai