arXivNur A Zarin Nishat, Jens Lehmann, Andrei Aioanei, Sahar VahdatiTue, Oct 6, 2026, 10:01 AM PDT
score 16.7
Small AI models memorize abstract-reasoning puzzles rather than learn rules
Original: A Systematic Study of Small Language Models on Abstract Reasoning Tasks
Source: arxiv.org ↗
Writing ELI5 summary…