Is One Layer Enough? Training A Single Transformer Layer Can Match Full-Parameter RL Training

• URL: http://arxiv.org/abs/2607.01232v2
• Date: Thu, 02 Jul 2026 17:28:23 GMT
• Title: Is One Layer Enough? Training A Single Transformer Layer Can Match Full-Parameter RL Training
• Abstract summary: Reinforcement learning (RL) has become a central component of post-training large language models.<n>We study how RL adaptation is distributed across transformer layers.<n>We observe a remarkably stable pattern: RL gains are highly concentrated in a small subset, and in many cases even a single, transformer layers.
• Score: 17.723489440201277
• License: http://creativecommons.org/licenses/by-nc-nd/4.0/
• Abstract: Reinforcement learning (RL) has become a central component of post-training large language models (LLMs), yet little is understood about how RL adaptation is distributed across transformer layers. Existing approaches typically update all model parameters uniformly, implicitly assuming that every layer contributes similarly to the gains obtained during RL post-training. In this work, we challenge this assumption through a systematic layer-wise study of RL training. Surprisingly, we find that training a single transformer layer can recover most of the gains achieved by full-parameter RL training, and in some cases even surpass it. To quantify this phenomenon, we introduce the quantity layer contribution, which measures the fraction of full RL improvement recovered by training a layer in isolation. Across seven models spanning two model families (Qwen3, Qwen2.5), three RL algorithms (GRPO, GiGPO, Dr. GRPO), and multiple task domains including mathematical reasoning, code generation, and agentic decision-making, we observe a remarkably stable pattern: RL gains are highly concentrated in a small subset of, and in many cases even a single, transformer layers. More strikingly, the same structural pattern consistently emerges: high-contribution layers concentrate in the middle of the transformer stack, while layers near the input and output ends contribute substantially less. The resulting layer rankings remain strongly correlated across datasets, tasks, model families, and RL algorithms.

Related papers

• RL-AD-Net: Reinforcement Learning Guided Adaptive Displacement in Latent Space for Refined Point Cloud Completion [9.252819624397405]
  We propose RL-AD-Net, a reinforcement learning framework that operates in the latent space of a pretrained point autoencoder.<n>To ensure robustness, a lightweight non-parametric PointNN selector evaluates the geometric consistency of both the original completion and the RL-refined output.<n>Experiments on ShapeNetCore-2048 demonstrate that while baseline completion networks perform reasonable under their training-style cropping, they struggle in random cropping scenarios.
  arXiv  Detail & Related papers  (2025-11-21T08:55:55Z)
• Reinforcement Learning Fine-Tunes a Sparse Subnetwork in Large Language Models [0.0]
  It is often assumed thatReinforcement learning (RL) fine-tuning requires updating most of a model's parameters.<n>We call this phenomenon RL-induced parameter update sparsity.<n>We show that fine-tuning only this sparse subnetwork recovers full model performance and yields parameters nearly identical to the fully fine-tuned model.
  arXiv  Detail & Related papers  (2025-07-23T01:02:17Z)
• 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)
• Scaling Offline RL via Efficient and Expressive Shortcut Models [13.050231036248338]
  offline reinforcement learning (RL) remains challenging due to the iterative nature of their noise sampling processes.<n>We introduce Scalable Offline Reinforcement Learning (SORL), a new offline RL algorithm that leverages shortcut models to scale both training and inference.<n>We demonstrate that SORL achieves strong performance across a range of offline RL tasks and exhibits positive scaling behavior with increased test-time compute.
  arXiv  Detail & Related papers  (2025-05-28T20:59:22Z)
• RAGEN: Understanding Self-Evolution in LLM Agents via Multi-Turn Reinforcement Learning [125.96848846966087]
  Training large language models (LLMs) as interactive agents presents unique challenges.<n>While reinforcement learning has enabled progress in static tasks, multi-turn agent RL training remains underexplored.<n>We propose StarPO, a general framework for trajectory-level agent RL, and introduce RAGEN, a modular system for training and evaluating LLM agents.
  arXiv  Detail & Related papers  (2025-04-24T17:57:08Z)
• StreamRL: Scalable, Heterogeneous, and Elastic RL for LLMs with Disaggregated Stream Generation [55.75008325187133]
  Reinforcement learning (RL) has become the core post-training technique for large language models (LLMs)<n>StreamRL is designed with disaggregation from first principles to address two types of performance bottlenecks.<n> Experiments show that StreamRL improves throughput by up to 2.66x compared to existing state-of-the-art systems.
  arXiv  Detail & Related papers  (2025-04-22T14:19:06Z)
• 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)
• ArCHer: Training Language Model Agents via Hierarchical Multi-Turn RL [80.10358123795946]
  We develop a framework for building multi-turn RL algorithms for fine-tuning large language models. Our framework adopts a hierarchical RL approach and runs two RL algorithms in parallel. Empirically, we find that ArCHer significantly improves efficiency and performance on agent tasks.
  arXiv  Detail & Related papers  (2024-02-29T18:45:56Z)
• PRILoRA: Pruned and Rank-Increasing Low-Rank Adaptation [65.268245109828]
  We introduce PRILoRA, which linearly allocates a different rank for each layer, in an increasing manner, and performs pruning throughout the training process. We validate the effectiveness of PRILoRA through extensive experiments on eight GLUE benchmarks, setting a new state of the art.
  arXiv  Detail & Related papers  (2024-01-20T20:25:17Z)
• Supplementing Gradient-Based Reinforcement Learning with Simple Evolutionary Ideas [4.873362301533824]
  We present a simple, sample-efficient algorithm for introducing large but directed learning steps in reinforcement learning (RL) The methodology uses a population of RL agents training with a common experience buffer, with occasional crossovers and mutations of the agents in order to search efficiently through the policy space.
  arXiv  Detail & Related papers  (2023-05-10T09:46:53Z)

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.