Evaluation & Benchmarks · 2025
ARC-AGI-2: A New Challenge for Frontier AI Reasoning Systems
A harder second-generation abstract-reasoning benchmark that stays easy for humans but collapses frontier models, keeping a measurable gap on fluid intelligence rather than memorized knowledge.
Editorial record
Plain-language summary
A harder second-generation abstract-reasoning benchmark that stays easy for humans but collapses frontier models, keeping a measurable gap on fluid intelligence rather than memorized knowledge.
Knowledge graph
Relationships
Descendants
Conceptual ancestorEvidence: Strongly supported
MMLU: Massive Multitask Language Understanding
Builds on the evaluation lineage in the archive
freshness sweep 2026
Source record
Provenance
- Record ID
- P-621
- Record created
- 2026-07-13
- Last reviewed
- 2026-07-14
- Record version
- 2
- https://arxiv.org/abs/2505.11831
- arXiv:2505.11831
Citation caveat: Citation metadata is approximate and marked unverified in the source dataset.