DyJR: Preserving Diversity in Reinforcement Learning with Verifiable Rewards via Dynamic Jensen-Shannon Replay

arXiv:2603.16157 · cs.LG, cs.AI · Submitted 2026-03-17 · Read on arXiv

cs.LG, cs.AI

Submitted: 2026-03-17

Updated: 2026-10-12

Comments: Accepted at NeurIPS 2026. 30 pages, 4 figures

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Terminology

• GRPO — 4× in this paper · explained in 6 episodes
• Regularization — 2× in this paper · explained in 3 episodes
• Reinforcement Learning — 2× in this paper · explained in 4 episodes
• Accuracy — 1× in this paper · explained in 1 episode
• Benchmark — 1× in this paper · explained in 2 episodes
• Mode Collapse — 1× in this paper · explained in 2 episodes
• Reinforcement Learning (RL) — 1× in this paper · explained in 25 episodes
• Text-to-SQL — 1× in this paper · explained in 3 episodes
• Verifiable Rewards — 1× in this paper · explained in 2 episodes

Sources

• VI-CuRL: Stabilizing Verifier-Independent RL Reasoning via Confidence-Guided Variance Reduction
• Reinforcement Learning with Verifiable yet Noisy Rewards under Imperfect Verifiers
• Towards Reasoning Era: A Survey of Long Chain-of-Thought for Reasoning Large Language Models
• Reasoning with Exploration: An Entropy Perspective
• Unlocking Reasoning Capabilities in LLMs via Reinforcement Learning Exploration
• AReaL: A Large-Scale Asynchronous Reinforcement Learning System for Language Reasoning
• DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning
• Tulu 3: Pushing Frontiers in Open Language Model Post-Training
• The Choice of Divergence: A Neglected Key to Mitigating Diversity Collapse in Reinforcement Learning with Verifiable Reward
• RePO: Replay-Enhanced Policy Optimization
• Squeeze the Soaked Sponge: Efficient Off-policy Reinforcement Finetuning for Large Language Model
• Continuous control with deep reinforcement learning
• Automated Optimization Modeling via a Localizable Error-Driven Perspective
• ARPO:End-to-End Policy Optimization for GUI Agents with Experience Replay
• Playing Atari with Deep Reinforcement Learning
• GPT-4 Technical Report
• Beyond the Sampled Token: Preserving Candidate Support in RLVR
• DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models
• Bottom-up Policy Optimization: Your Language Model Policy Secretly Contains Internal Policies
• Kimi k1.5: Scaling Reinforcement Learning with LLMs

Related papers

• Polynomial-Augmented Neural Networks (PANNs) with Weak Orthogonality Constraints for Enhanced Function and PDE Approximation
• AIRL-S: Unifying Reinforcement Learning and Search-Based Test-Time Scaling via Adversarial Inverse Reinforcement Learning
• Transformers as Bayesian In-Context Experimenters: Smoothness-Adaptive Efficient ATE Estimation
• Convergence issues in Relational Concept Analysis based on AOC-posets
• Beliefs Beyond Posteriors: The Predictive Cost of ELBO Binding
• Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks