ARC-AGI-2
ARC-AGI-2 evaluates few-shot abstraction, rule induction, and exact generalization on unfamiliar visual grid transformations under the competition's resource constraints.
ARC-AGI-2 ARC-AGI-2 2025 private evaluation
Permanent entry- Task and data
- ARC-AGI-2 evaluates programmatic solutions on novel abstract grid-transformation tasks; the 2025 competition used a hidden private evaluation under efficiency and cost constraints.
- Primary metric
- A task scores as solved when one of two submitted outputs exactly matches the hidden output; the final accuracy is averaged over private tasks and higher is better.
- What it measures
- few-shot abstraction, rule induction, and exact generalization on unfamiliar visual grid transformations under the competition's resource constraints.
- Comparison rules
- Compare only the same ARC-AGI-2 year, split, submission allowance, and budget. These are 2025 competition system results, not model-only scores.
- Access and licensing
- Use ARC data and competition assets only under their official terms. The Compendium reproduces no grids, tasks, or solution artifacts.
| Published subject | Score | Metric | Split and evaluation setting | Source record |
|---|---|---|---|---|
| NVARCarc-agi-2:2025:private:budget-0.20 | 24.0%higher, percent | private-evaluation accuracyofficial 2025 competition result | ARC-AGI-2 2025 private evaluationOfficial competition result at a reported cost of USD $0.20 per task; implementation details live in the team's linked write-up.This is an engineered competition solution, not a base-model evaluation. | ARC Prize FoundationPublished not stated by publisher, reviewed 2026-07-13competition-result, published-third-party |
| the ARChitectsarc-agi-2:2025:private:budget-0.20 | 16.5%higher, percent | private-evaluation accuracyofficial 2025 competition result | ARC-AGI-2 2025 private evaluationOfficial competition result at a reported cost of USD $0.20 per task; implementation details live in the team's linked write-up.The aggregate page does not restate every inference and training parameter; no missing setup is inferred here. | ARC Prize FoundationPublished not stated by publisher, reviewed 2026-07-13competition-result, published-third-party |
| MindsAIarc-agi-2:2025:private:budget-0.20 | 12.6%higher, percent | private-evaluation accuracyofficial 2025 competition result | ARC-AGI-2 2025 private evaluationOfficial competition result at a reported cost of USD $0.20 per task; implementation details live in the team's linked write-up.Later ARC-AGI-2 leaderboards and different budgets are separate comparison classes. | ARC Prize FoundationPublished not stated by publisher, reviewed 2026-07-13competition-result, published-third-party |
Claims and evidence
Claim ledger, 2ARC-AGI-2 evaluates programmatic solutions on novel abstract grid-transformation tasks; the 2025 competition used a hidden private evaluation under efficiency and cost constraints.
The listed result anchors are third-party publications tied to explicit evaluation settings and are not evaluations run by Draconitic.