Grid-based reasoning benchmark testing how efficiently AI systems infer unfamiliar rules from a small number of examples.

Interactive reasoning benchmark measuring how efficiently AI agents explore novel environments, learn goals, and adapt through experience.

Independent benchmark verification program evaluating AI systems on hidden ARC-AGI datasets, with transparent methodology and academic oversight.

Human-calibrated static reasoning benchmark testing symbolic interpretation, compositional reasoning, and efficient adaptation to unfamiliar tasks.

Open-source research competition rewarding progress on static and interactive ARC-AGI benchmarks, alongside papers advancing general intelligence.