Internalizing Meta-Experience into Memory for Guided Reinforcement Learning in Large Language Models

TL;DR

This paper introduces Meta-Experience Learning (MEL), a framework that internalizes error-based knowledge into language models' memory to improve reasoning and performance in reinforcement learning tasks.

Contribution

MEL is a novel framework that leverages self-verification and contrastive analysis to internalize meta-experience into LLMs, enhancing their reasoning and learning capabilities.

Findings

Achieves 3.92%--4.73% Pass@1 improvements across benchmarks.

Effectively internalizes error analysis into model memory.

Enhances fine-grained credit assignment in reinforcement learning.

Abstract

Reinforcement Learning with Verifiable Rewards (RLVR) has emerged as an effective approach for enhancing the reasoning capabilities of Large Language Models (LLMs). Despite its efficacy, RLVR faces a meta-learning bottleneck: it lacks mechanisms for error attribution and experience internalization intrinsic to the human learning cycle beyond practice and verification, thereby limiting fine-grained credit assignment and reusable knowledge formation. We term such reusable knowledge representations derived from past errors as meta-experience. Based on this insight, we propose Meta-Experience Learning (MEL), a novel framework that incorporates self-distilled meta-experience into the model's parametric memory. Building upon standard RLVR, we introduce an additional design that leverages the LLM's self-verification capability to conduct contrastive analysis on paired correct and incorrect trajectories, identify the precise bifurcation points where reasoning errors arise, and summarize them into generalizable meta-experience. The meta-experience is further internalized into the LLM's parametric memory by minimizing the negative log-likelihood, which induces a language-modeled reward signal that bridges correct and incorrect reasoning trajectories and facilitates effective knowledge reuse. Experimental results demonstrate that MEL achieves consistent improvements on benchmarks, yielding 3.92%--4.73% Pass@1 gains across varying model sizes.