
https://arcprize.org/leaderboard
Achieve a performance of 85% or higher on the private set of Chollet's Abstraction and Reasoning Corpus (ARC), a challenging benchmark designed to test an AI's ability to solve abstract reasoning tasks with minimal prior knowledge. Even if the requirements for the grand prize change in 2025, this benchmark of >=85% performance will remain unchanged. Learn more about the competition here: https://arcprize.org/competition.
2025 version https://manifold.markets/JacobPfau/will-the-arcagi-grand-prize-be-clai-srb6t2awj1
2024 version https://manifold.markets/JacobPfau/will-the-arcagi-grand-prize-be-clai
People are also trading
Is this on ARC-AGI-1? GPT-5.2 Pro (High) is already >=85%: https://arcprize.org/leaderboard
@adssx Ah it's on the private set... So we run into that issue: https://manifold.markets/JacobPfau/will-the-arcagi-grand-prize-be-clai-srb6t2awj1#062h5wqq298j
@louuse What are the resolution criteria for this? I think it needs to resolve either YES (as multiple AIs now have >85% on ARC-AGI-1) or N/A (as we don't have data specifically on the "private set", only "verified scores" / "semi-private set" from my understanding).
