A Systematic Study of Small Language Models on Abstract Reasoning Tasks

The paper investigates how small language models acquire abstract reasoning skills on the ARC‑TGI benchmark, which groups grid‑transformation tasks into controllable families and allows resampling, spatial shifts, and cross‑benchmark transfer. Over 1,000 supervised fine‑tuning runs across decoder‑only, encoder‑decoder, and mixture‑of‑experts families, the study finds that high in‑distribution accuracy is possible but depends heavily on optimization and is uneven across task families. Performance drops sharply outside the training distribution, and gains from larger training sets or additional in‑context examples vary by model family; attention diagnostics reveal distinct patterns but do not explain causal mechanisms.

Machine-generated by The Flow from the publisher's headline and feed description — not written or checked by a human. The full article lives at arXiv Machine Learning.

• llms
• fine-tuning
• benchmarks