The Choice of Divergence: A Neglected Key to Mitigating Diversity Collapse in Reinforcement Learning with Verifiable Reward

• URL: http://arxiv.org/abs/2509.07430v2
• Date: Fri, 17 Oct 2025 09:26:39 GMT
• Title: The Choice of Divergence: A Neglected Key to Mitigating Diversity Collapse in Reinforcement Learning with Verifiable Reward
• Authors: Long Li, Jiaran Hao, Jason Klein Liu, Zhijian Zhou, Yanting Miao, Wei Pang, Xiaoyu Tan, Wei Chu, Zhe Wang, Shirui Pan, Chao Qu, Yuan Qi,
• Abstract summary: A central paradox in fine-tuning Large Language Models (LLMs) with Reinforcement Learning with Verifiable Reward (RLVR) is the frequent degradation of multi-attempt performance.<n>This is often accompanied by catastrophic forgetting, where models lose previously acquired skills.<n>We argue that standard RLVR objectives lack a crucial mechanism for knowledge retention.
• Score: 57.56453588632619
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: A central paradox in fine-tuning Large Language Models (LLMs) with Reinforcement Learning with Verifiable Reward (RLVR) is the frequent degradation of multi-attempt performance (Pass@k) despite improvements in single-attempt accuracy (Pass@1). This is often accompanied by catastrophic forgetting, where models lose previously acquired skills. While various methods have been proposed, the choice and function of the divergence term have been surprisingly unexamined as a proactive solution. We argue that standard RLVR objectives -- both those using the mode-seeking reverse KL-divergence and those forgoing a divergence term entirely -- lack a crucial mechanism for knowledge retention. The reverse-KL actively accelerates this decay by narrowing the policy, while its absence provides no safeguard against the model drifting from its diverse knowledge base. We propose a fundamental shift in perspective: using the divergence term itself as the solution. Our framework, Diversity-Preserving Hybrid RL (DPH-RL), leverages mass-covering f-divergences (like forward-KL and JS-divergence) to function as a rehearsal mechanism. By continuously referencing the initial policy, this approach forces the model to maintain broad solution coverage. Extensive experiments on math and SQL generation demonstrate that DPH-RL not only resolves the Pass@k degradation but improves both Pass@1 and Pass@k in- and out-of-domain. Additionally, DPH-RL is more training-efficient because it computes f-divergence using generator functions, requiring only sampling from the initial policy and no online reference model. Our work highlights a crucial, overlooked axis for improving RLVR, demonstrating that the proper selection of a divergence measure is a powerful tool for building more general and diverse reasoning models.

Related papers

• Controllable Exploration in Hybrid-Policy RLVR for Multi-Modal Reasoning [88.42566960813438]
  CalibRL is a hybrid-policy RLVR framework that supports controllable exploration with expert guidance.<n>CalibRL increases policy entropy in a guided manner and clarifies the target distribution.<n>Experiments across eight benchmarks, including both in-domain and out-of-domain settings, demonstrate consistent improvements.
  arXiv  Detail & Related papers  (2026-02-22T07:23:36Z)
• Data-regularized Reinforcement Learning for Diffusion Models at Scale [99.01056178660538]
  We introduce Data-regularized Diffusion Reinforcement Learning ( DDRL), a novel framework that uses the forward KL divergence to anchor the policy to an off-policy data distribution.<n>With over a million GPU hours of experiments and ten thousand double-blind evaluations, we demonstrate that DDRL significantly improves rewards while alleviating the reward hacking seen in RLs.
  arXiv  Detail & Related papers  (2025-12-03T23:45:07Z)
• GRPO Privacy Is at Risk: A Membership Inference Attack Against Reinforcement Learning With Verifiable Rewards [13.369116707284121]
  Divergence-in-Behavior Attack (DIBA) is the first membership inference framework specifically designed for Reinforcement Learning with Verifiable Rewards.<n>We show that DIBA significantly outperforms existing baselines, achieving around 0.8 AUC and an order-of-magnitude higher TPR@0.1%FPR.<n>This is the first work to systematically analyze privacy vulnerabilities in RLVR, revealing that training data exposure can be reliably inferred through behavioral traces.
  arXiv  Detail & Related papers  (2025-11-18T01:51:34Z)
• HINT: Helping Ineffective Rollouts Navigate Towards Effectiveness [49.72591739116668]
  Reinforcement Learning (RL) has become a key driver for enhancing the long chain-of-thought (CoT) reasoning capabilities of Large Language Models (LLMs)<n>However, prevalent methods like GRPO often fail when task difficulty exceeds the model's capacity, leading to reward sparsity and inefficient training.<n>We propose HINT: Helping Ineffective rollouts Navigate Towards effectiveness, an adaptive hinting framework.
  arXiv  Detail & Related papers  (2025-10-10T13:42:03Z)
• The Reasoning Boundary Paradox: How Reinforcement Learning Constrains Language Models [31.773914661815393]
  Reinforcement Learning with Verifiable Rewards (RLVR) has emerged as a key method for improving Large Language Models' reasoning capabilities.<n>Recent evidence suggests it may paradoxically shrink the reasoning boundary rather than expand it.<n>This paper investigates the shrinkage issue of RLVR by analyzing its learning dynamics.
  arXiv  Detail & Related papers  (2025-10-02T17:17:27Z)
• MMR1: Enhancing Multimodal Reasoning with Variance-Aware Sampling and Open Resources [113.33902847941941]
  Variance-Aware Sampling (VAS) is a data selection strategy guided by Variance Promotion Score (VPS)<n>We release large-scale, carefully curated resources containing 1.6M long CoT cold-start data and 15k RL QA pairs.<n> Experiments across mathematical reasoning benchmarks demonstrate the effectiveness of both the curated data and the proposed VAS.
  arXiv  Detail & Related papers  (2025-09-25T14:58:29Z)
• RL as Regressor: A Reinforcement Learning Approach for Function Approximation [0.0]
  We propose framing regression as a Reinforcement Learning (RL) problem.<n>We demonstrate this by treating a model's prediction as an action and defining a custom reward signal based on the prediction error.<n>We show that the RL framework not only successfully solves the regression problem but also offers enhanced flexibility in defining objectives and guiding the learning process.
  arXiv  Detail & Related papers  (2025-07-31T21:39:24Z)
• The Invisible Leash: Why RLVR May Not Escape Its Origin [48.915013455847856]
  Recent advances in large reasoning models highlight Reinforcement Learning with Verifiable Rewards (RLVR) as a promising method for enhancing AI's capabilities.<n>This study presents a theoretical and empirical investigation that provides fresh insights into the potential limits of RLVR.<n>We identify an entropy-reward tradeoff: while RLVR reliably enhances precision, it may progressively narrow exploration and potentially overlook correct yet underrepresented solutions.
  arXiv  Detail & Related papers  (2025-07-20T07:04:08Z)
• PRDP: Proximal Reward Difference Prediction for Large-Scale Reward Finetuning of Diffusion Models [13.313186665410486]
  Reward finetuning has emerged as a promising approach to aligning foundation models with downstream objectives. Existing reward finetuning methods are limited by their instability in large-scale prompt datasets. We propose Proximal Reward Difference Prediction (PRDP) to enable stable black-box reward finetuning for diffusion models.
  arXiv  Detail & Related papers  (2024-02-13T18:58:16Z)
• Deep Generative Symbolic Regression [83.04219479605801]
  Symbolic regression aims to discover concise closed-form mathematical equations from data. Existing methods, ranging from search to reinforcement learning, fail to scale with the number of input variables. We propose an instantiation of our framework, Deep Generative Symbolic Regression.
  arXiv  Detail & Related papers  (2023-12-30T17:05:31Z)
• REBEL: Reward Regularization-Based Approach for Robotic Reinforcement Learning from Human Feedback [61.54791065013767]
  A misalignment between the reward function and human preferences can lead to catastrophic outcomes in the real world.<n>Recent methods aim to mitigate misalignment by learning reward functions from human preferences.<n>We propose a novel concept of reward regularization within the robotic RLHF framework.
  arXiv  Detail & Related papers  (2023-12-22T04:56:37Z)
• Understanding, Predicting and Better Resolving Q-Value Divergence in Offline-RL [86.0987896274354]
  We first identify a fundamental pattern, self-excitation, as the primary cause of Q-value estimation divergence in offline RL. We then propose a novel Self-Excite Eigenvalue Measure (SEEM) metric to measure the evolving property of Q-network at training. For the first time, our theory can reliably decide whether the training will diverge at an early stage.
  arXiv  Detail & Related papers  (2023-10-06T17:57:44Z)

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.