Why Does Reinforcement Learning Generalize? A Feature-Level Mechanistic Study of Post-Training in Large Language Models

• URL: http://arxiv.org/abs/2604.25011v1
• Date: Mon, 27 Apr 2026 21:22:34 GMT
• Title: Why Does Reinforcement Learning Generalize? A Feature-Level Mechanistic Study of Post-Training in Large Language Models
• Abstract summary: Reinforcement learning (RL)-based post-training often improves the reasoning performance of large language models beyond the training domain.<n>Supervised fine-tuning (SFT) frequently leads to general forgetting capabilities.<n>We present a feature-level mechanistic analysis methodology to probe RL generalization using a controlled experimental setup.
• Score: 51.45122910254346
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Reinforcement learning (RL)-based post-training often improves the reasoning performance of large language models (LLMs) beyond the training domain, while supervised fine-tuning (SFT) frequently leads to general capabilities forgetting. However, the mechanisms underlying this contrast remain unclear. To bridge this gap, we present a feature-level mechanistic analysis methodology to probe RL generalization using a controlled experimental setup, where RL- and SFT-tuned models are trained from the same base model on identical data. Leveraging our interpretability framework, we align internal activations across models within a shared feature space and analyze how features evolve during post-training. We find that SFT rapidly introduces many highly specialized features that stabilize early in training, whereas RL induces more restrained and continually evolving feature changes that largely preserve base models' representations. Focusing on samples where RL succeeds but the base model fails, we identify a compact, task-agnostic set of features that directly mediate generalization across diverse tasks. Feature-level interventions confirm their causal role: disabling these features significantly degrades RL models' generalization performance, while amplifying them improves base models' performance. The code is available at https://github.com/danshi777/RL-generalization.

Related papers

• What Does an LLM Learn from Reinforcement Learning? A Mechanistic Interpretability Perspective with Fixed-SAE Track [4.624097215050827]
  We introduce Fixed-SAE Track, a framework that trains one shared SAE per layer on activations pooled across the base model and all RL checkpoints.<n>We find that RL-induced drift is small, gradual, concept specific, and concentrated in late layers, mainly enhancing the sampling rates of a small set of ladder tokens.
  arXiv  Detail & Related papers  (2026-09-14T05:29:45Z)
• Learning Dynamics in RL Post-Training for Language Models [2.538209532048867]
  We analyze the learning dynamics of RL post-training from a perspective that has been studied in supervised learning but remains underexplored in RL.<n>We show that limited variability in feature representations can cause RL updates to systematically increase model confidence.<n>Motivated by these insights, we propose classifier-first reinforcement learning (CF-RL), a simple two-stage training strategy.
  arXiv  Detail & Related papers  (2026-01-08T07:32:15Z)
• Nemotron-Cascade: Scaling Cascaded Reinforcement Learning for General-Purpose Reasoning Models [71.9060068259379]
  We propose cascaded domain-wise reinforcement learning to build general-purpose reasoning models.<n>Our 14B model, after RL, outperforms its SFT teacher, DeepSeek-R1-0528, on LiveCodeBench v5/v6 Pro and silver-medal performance in the 2025 International Olympiad in Informatics (IOI)
  arXiv  Detail & Related papers  (2025-12-15T18:02:35Z)
• Tool Zero: Training Tool-Augmented LLMs via Pure RL from Scratch [63.40752011615843]
  Training tool-augmented language models has emerged as a promising approach to enhancing their capabilities for complex tasks.<n>We propose a dynamic generalization-guided reward design for rule-based reinforcement learning.<n>We show that our models achieve over 7% performance improvement compared to both SFT and RL-with-SFT models.
  arXiv  Detail & Related papers  (2025-11-02T16:33:45Z)
• Beyond Accuracy: Dissecting Mathematical Reasoning for LLMs Under Reinforcement Learning [93.00629872970364]
  Reinforcement learning (RL) has become the dominant paradigm for improving the performance of language models on complex reasoning tasks.<n>We introduce SPARKLE, a fine-grained analytic framework to dissect the effects of RL across three key dimensions.<n>We study whether difficult problems -- those yielding no RL signals and mixed-quality reasoning traces -- can still be effectively used for training.
  arXiv  Detail & Related papers  (2025-06-05T07:53:59Z)
• Echo Chamber: RL Post-training Amplifies Behaviors Learned in Pretraining [74.83412846804977]
  Reinforcement learning (RL)-based fine-tuning has become a crucial step in post-training language models.<n>We present a systematic end-to-end study of RL fine-tuning for mathematical reasoning by training models entirely from scratch.
  arXiv  Detail & Related papers  (2025-04-10T17:15:53Z)
• SFT Memorizes, RL Generalizes: A Comparative Study of Foundation Model Post-training [127.47044960572659]
  Supervised fine-tuning (SFT) and reinforcement learning (RL) are widely used post-training techniques for foundation models.<n>This paper studies the difference between SFT and RL on generalization and memorization.<n>We show that RL, especially when trained with an outcome-based reward, generalizes across both rule-based textual and visual variants.
  arXiv  Detail & Related papers  (2025-01-28T18:59:44Z)
• Improving Zero-shot Generalization in Offline Reinforcement Learning using Generalized Similarity Functions [34.843526573355746]
  Reinforcement learning (RL) agents are widely used for solving complex sequential decision making tasks, but exhibit difficulty in generalizing to scenarios not seen during training. We show that performance of online algorithms for generalization in RL can be hindered in the offline setting due to poor estimation of similarity between observations. We propose a new theoretically-motivated framework called Generalized Similarity Functions (GSF), which uses contrastive learning to train an offline RL agent to aggregate observations based on the similarity of their expected future behavior.
  arXiv  Detail & Related papers  (2021-11-29T15:42:54Z)

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.