Evidence Passport
ARCANA: A Benchmark for Abstraction and Analogical Reasoning
Top story 7 in the issue of 23 October 2026, in 48 AI AGI & ASI.
Original source
ARCANA: A Benchmark for Abstraction and Analogical Reasoning
ARCANA asks solvers to map real objects and physical effects onto abstract grid puzzles; people do well, while language models fall clearly short.
Evidence
Evidence: Preprint (evidence step 4 of 6)
A research paper shared publicly before peer review.
The story in three lines
- Key finding
- People reliably solve ARCANA's far-analogy puzzles, while a range of large language models show a clear performance gap.
- Significance
- Technical: transferring ideas between distant domains remains a weak point for current models.
- Takeaway
- A useful new yardstick for general reasoning; results are not yet peer-reviewed.
Later coverage in the journal
No later coverage in the journal yet.
Corrections and updates
No corrections or updates recorded for this item.
This page is generated from the issue and stays at the same address. It adds nothing that is not in the issue, the source link or the corrections log.