From Reasoning Traces to Reusable Modules: Understanding Compositional Generalization in Language Model Reasoning

• URL: http://arxiv.org/abs/2606.18089v1
• Date: Tue, 16 Jun 2026 15:55:28 GMT
• Title: From Reasoning Traces to Reusable Modules: Understanding Compositional Generalization in Language Model Reasoning
• Abstract summary: Post-training pipelines combine supervised fine-tuning with reinforcement learning.<n>We show that SFT supplies the raw module materials in compositional traces, and RL decomposes those traces to identify the latent atomic modules.<n>We find that training on compound traces yields stronger generalization than training on isolated atomic modules.
• Score: 92.21289223393067
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Post-training pipelines that combine supervised fine-tuning (SFT) with reinforcement learning (RL) have emerged as the key recipe for transforming large language models (LLMs) into robust reasoners. We argue that this combined success is driven by compositional generalization, which we formalize through a hierarchical latent selection model. In this framework, reasoning traces are generated by a cascade of discrete latent selection variables corresponding to reusable atomic modules, including both skills (local operations) and routing mechanisms (how intermediate information is selected, reused, and composed). Within this model, we theoretically show that SFT and RL play asymmetric, complementary roles: SFT supplies the raw module materials in compositional traces, and RL decomposes those traces to identify the latent atomic modules and enable compositional generalization. We design controlled experiments to validate this theory. Our results demonstrate that RL can extract atomic modules from compound traces supplied by SFT and recombine them to solve new configurations. Moreover, we find that training on compound traces yields stronger generalization than training on isolated atomic modules. Finally, we investigate the relationship between SFT and RL data and identify an effective protocol in which SFT ensures coverage of all atomic modules through compositional traces, while RL focuses on novel compositions outside the SFT support to drive exploration.

Related papers

• From Global to Factor-Wise Expert Composition in Discrete Diffusion Models [5.756282731737539]
  We propose FactorDiff - a factor-wise composition framework for diffusion models.<n>We instantiate this framework with spatial/pixel-level compositions and validate it on the ARC-AGI benchmark.
  arXiv  Detail & Related papers  (2026-07-13T16:22:01Z)
• Learning What to Learn: Stage-Specific Data Sets for SFT-then-RL in Small Language Model Reasoning [29.726247369977198]
  We argue that data strategy should be aligned with the distinct roles of SFT and RL.<n>We propose a difficulty-aware SFT-then-RL framework that organizes training data into stage-specific sets.
  arXiv  Detail & Related papers  (2026-06-03T05:25:06Z)
• Entropy-KL Divergence-based Token Masking: A Novel Approach for Selective Fine-tuning of Large Language Models [52.11240605311707]
  Supervised fine-tuning followed by reinforcement learning has become a standard post-training paradigm for large language models.<n>We propose EKSFT (Entropy-KL Selective Fine-Tuning), which selectively masks tokens that exhibit either high entropy or high KL divergence from a reference model.<n> Empirical evaluations on mathematical reasoning benchmarks demonstrate that EKSFT consistently outperforms standard SFT.
  arXiv  Detail & Related papers  (2026-05-28T03:36:05Z)
• Influence-Guided Symbolic Regression: Scientific Discovery via LLM-Driven Equation Search with Granular Feedback [56.69850045068714]
  We introduce textitInfluence-Guided Regression (IGSR), a method that frames equation discovery as an iterative two-step process.<n>We demonstrate IGSRs effectiveness on a diverse suite of benchmarks, including LLM-SRBench, pharmacological PKPD models, an epidemiological simulation, and real-world genomic data.
  arXiv  Detail & Related papers  (2026-05-27T23:48:01Z)
• Consolidation or Adaptation? PRISM: Disentangling SFT and RL Data via Gradient Concentration [56.074760766965085]
  PRISM achieves a dynamics-aware framework that arbitrates data based on its degree of cognitive conflict with the model's existing knowledge.<n>Our findings suggest that disentangling data based on internal optimization regimes is crucial for scalable and robust agent alignment.
  arXiv  Detail & Related papers  (2026-01-12T05:43:20Z)
• From Atomic to Composite: Reinforcement Learning Enables Generalization in Complementary Reasoning [83.94543243783285]
  We study Complementary Reasoning, a complex task that requires integrating internal parametric knowledge with external contextual information.<n>We find that RL acts as a reasoning synthesizer rather than a probability amplifier.
  arXiv  Detail & Related papers  (2025-12-01T18:27:25Z)
• Round-trip Reinforcement Learning: Self-Consistent Training for Better Chemical LLMs [51.29260537017623]
  Large Language Models (LLMs) are emerging as versatile foundation models for computational chemistry.<n>These models often lack round-trip consistency.<n>We introduce Round-Trip Reinforcement Learning (RTRL), a novel framework that trains a model to improve its consistency.
  arXiv  Detail & Related papers  (2025-10-01T23:58:58Z)
• Unlock the Correlation between Supervised Fine-Tuning and Reinforcement Learning in Training Code Large Language Models [12.656574142412484]
  We make an attempt to understand the correlation between supervised fine-tuning and reinforcement learning.<n>We find that both atomic and synthetic functions are indispensable for SFT's generalization.
  arXiv  Detail & Related papers  (2024-06-14T03:39:01Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.