Notes to Self: Can LLMs Benefit from Experiential Abstractions?
2026-07-22 • Computation and Language
Computation and Language
AI summaryⓘ
The authors studied whether Large Language Models (LLMs) can learn useful general ideas, like humans do, from their problem-solving experiences. They had stronger models or the LLMs themselves create a library of simple explanations called abstractions from math problems. These abstractions were then used during problem-solving or to help train the models, improving their math and logic skills. The authors found that LLMs can create and use these abstractions similarly to how humans apply past lessons. This approach also works well across different datasets and models.
Large Language Modelsabstractionsinference-time retrievalreinforcement learningtraining promptsmathematical reasoninglogical reasoningself-extractiontransfer learning
Authors
Chang Liu, Xinyu Li, Artur Dubrawski
Abstract
Humans distill experience into reusable abstractions, e.g., strategies and cautionary reminders, and apply them to gradually solve problems more effectively. We study whether Large Language Models (LLMs) can similarly benefit from such experiential abstractions. From LLMs' solution traces on the MATH training set, a stronger teacher or the LLMs themselves extract natural-language abstractions into a retrievable library. We explore two usage modes: (1) inference-time retrieval and (2) reinforcement learning (RL) with abstraction-augmented training prompts. Experiential abstractions improve LLM performance on mathematical and logical reasoning benchmarks. Self-extracted abstractions match teacher-extracted ones, and our abstraction usage framework can transfer to other datasets and models. These findings suggest LLMs can extract and apply experiential abstractions much as humans leverage distilled experience.