ARC-AGI-3
ARC Prize / Reasoning · Learning and generalising from unfamiliar rules
What it measures, and how
The standard harness is fixed and kept separate from provider-specific optimised adapter results.
How this evidence is used
Under observation: the official site's Terms require written permission for systematic collection and aggregation; the aggregation boundary for current official scores is still pending. The standard harness and Provider Adapter are kept separate, and the top score of a dedicated adapter is not taken to represent an ordinary model.
Limits and data attribution
Under observation: the official site's Terms require written permission for systematic collection and aggregation; the aggregation boundary for current official scores is still pending. The standard harness and Provider Adapter are kept separate, and the top score of a dedicated adapter is not taken to represent an ordinary model.
Data licence: Leaderboard data usage boundaries to be confirmed
Scores are published by ARC Prize; the original scores and the News consensus score use different scales and cannot be added directly.