# A Systematic Study of Small Language Models on Abstract Reasoning Tasks

arXiv: https://arxiv.org/abs/2610.08680 · HTML: https://mlsanity.com/paper/2610.08680/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Test-time compute and efficient thinking, one of 17 directions in LLM reasoning.
- Builds on 17 of its 37 references that are in the map.
- First posted on arXiv 6 Oct 2026.

## Placed in

- LLM reasoning / [Test-time compute and efficient thinking](https://mlsanity.com/lr/test-time-compute-and-efficient-thinking/)

## Builds on (most cited first, 17 of 17)

- [Switch Transformers: Scaling to Trillion Parameter Models with Simple and Efficient Sparsity](https://mlsanity.com/paper/2101.03961/) (2021)
- [Grokking: Generalization Beyond Overfitting on Small Algorithmic Datasets](https://mlsanity.com/paper/2201.02177/) (2022)
- [ARC-AGI-2: A New Challenge for Frontier AI Reasoning Systems](https://mlsanity.com/paper/2505.11831/) (2025)
- [On the Measure of Intelligence](https://mlsanity.com/paper/1911.01547/) (2019)
- [Does Math Reasoning Improve General LLM Capabilities? Understanding Transferability of LLM Reasoning](https://mlsanity.com/paper/2507.00432/) (2025)
- [ARC Prize 2024: Technical Report](https://mlsanity.com/paper/2412.04604/) (2024)
- [The ConceptARC Benchmark: Evaluating Understanding and Generalization in the ARC Domain](https://mlsanity.com/paper/2305.07141/) (2023)
- [Large Language Models Are Not Strong Abstract Reasoners](https://mlsanity.com/paper/2305.19555/) (2023)
- [ARC-AGI-3: A New Challenge for Frontier Agentic Intelligence](https://mlsanity.com/paper/2603.24621/) (2026)
- [Addressing the Abstraction and Reasoning Corpus via Procedural Example Generation](https://mlsanity.com/paper/2404.07353/) (2024)
- [Encoder-Decoder Gemma: Improving the Quality-Efficiency Trade-Off via Adaptation](https://mlsanity.com/paper/2504.06225/) (2025)
- [Comment on The Illusion of Thinking: Understanding the Strengths and Limitations of Reasoning Models via the Lens of Problem Complexity](https://mlsanity.com/paper/2506.09250/) (2025)
- [ARC-GEN: A Mimetic Procedural Benchmark Generator for the Abstraction and Reasoning Corpus](https://mlsanity.com/paper/2511.00162/) (2025)
- [The ARC of Progress towards AGI: A Living Survey of Abstraction and Reasoning](https://mlsanity.com/paper/2603.13372/) (2026)
- [Reinforcement Learning Improves Traversal of Parametric Knowledge in LLMs](https://mlsanity.com/paper/2511.05933/) (2025)
- [GIFARC: Synthetic Dataset for Leveraging Human-Intuitive Analogies to Elevate AI Reasoning](https://mlsanity.com/paper/2505.20672/) (2025)
- [ARC-TGI: Human-Validated Task Generators with Reasoning Chain Templates for ARC-AGI](https://mlsanity.com/paper/2603.05099/) (2026)

## Abstract (arXiv)

Endpoint accuracy on abstract-reasoning benchmarks does not reveal whether a language model has acquired a transferable rule or fit distribution-specific regularities. We study this distinction in small language models on the ARC-TGI benchmark, which organizes abstract grid transformations into controllable task families and supports resampling, spatial shifts, and cross-benchmark transfer. Across more than 1,000 runs, we profile decoder-only, encoder--decoder, and mixture-of-experts model families under supervised fine-tuning. We examine the efficiency and stability of skill acquisition, robustness beyond the training distribution, interactions with model family and task formulation, and layer-wise attention signatures that accompany behavioral differences. Substantial in-distribution accuracy is attainable, but acquisition is sensitive to optimization and unevenly distributed across task families. Performance deteriorates sharply outside the training distribution, including when the rule is retained but grid scale changes. Greater training-set depth and breadth yield uneven gains, while the effect of additional in-context examples depends on model family. Executable-rule induction also yields correct solutions not observed under direct grid generation. On selected tasks, attention diagnostics show distinct concentration and context-dependence profiles, but do not establish general causal mechanisms. Overall, abstract-reasoning scores are conditional on the model, adaptation regime, evaluation distribution, and response format.