Four-Factor Integrity Breakdown & Primary-Source Evidence
Privacy
Weight 40%
7.5/10
Access Model: Hybrid Public / Private
"Public Eval Set 120 tasks Calibrated, public, all tasks solved pass@2 by at least two humans, use to test your systems. Semi-Private Eval Set 120 tasks Calibrated, not public, all tasks solved pass@2 by at least two humans, used for Kaggle live contest leaderboard and ARC Prize leaderboard."
"ARC-AGI-2 contains 1,000 public training tasks and 120 public evaluation tasks."
Originality
Weight 25%
10/10
Reuse Profile: All Original (0% reuse)
"The Abstraction and Reasoning Corpus for Artificial General Intelligence (ARC-AGI), introduced in 2019, established a challenging benchmark for evaluating the general fluid intelligence of artificial systems via a set of unique, novel tasks only requiring minimal prior knowledge."
Freshness
Weight 20%
6/10
Initial Release: 2025-03Latest Refresh: 2025-03Rolling Updates: No
"By Greg Kamradt Published 24 Mar 2025 ARC-AGI-2 + ARC Prize 2025 is Live!"
Integrity: Independent Offline (Leak-Verified)Sandbox: offline isolated
"Internet access is not available during Kaggle evaluation (no API-based systems like GPT/Claude/etc.)"
Frontier Model Standings on ARC-AGI-2
Top 15 active models tracked
Benchmark maintainers can embed this live SVG badge in GitHub READMEs or documentation:
[](https://benchaudit.org/benchmarks/arc-agi-2/)