Aug 2026· Proceedings of the 32nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining V.2· pp. 4046-4057· 0 citations· 4 references
Abstract
Reinforcement Learning with Verifiable Rewards (RLVR) has become a standard approach for improving complex reasoning in Large Language Models (LLMs). However, entropy collapse remains a critical bottleneck, often leading to performance degradation. While typically attributed to reduced exploration, we identify a more severe failure mode: update deadlock. We show that policy entropy bounds both the policy gradient and probability update norms; consequently, entropy collapse effectively stops reward signal backpropagation, preventing further policy learning regardless of data quality. To analyze this, we develop a micro-dynamics framework based on single-step logit updates. Our results indicate that entropy evolution depends on the interaction between advantage signals and the log-probability residual relative to its expectation, offering a unified perspective on existing interventions. Based on these findings, we propose ANCHOR, an algorithm designed to resolve deadlocks and maintain stability. ANCHOR uses a dual mechanism: it dynamically anchors the entropy decay of positive samples to prevent premature saturation, and uses entropy variations to stabilize updates for negative samples. Extensive experiments on reasoning benchmarks show that ANCHOR significantly outperforms strong baselines. Unlike heuristic methods that induce verbosity to mitigate collapse, ANCHOR promotes concise reasoning paths, improving both stability and efficiency.
A formal analysis showing that joint optimization of the two objectives induces gradient conflict in early training, motivating the sequential design of ERR+, a two-phase RLVR framework grounded in this observation.
Xinle Jiang, Min-Hao Wang, Wen Wu et al.· 0 citations
This work empirically study the potential of outer prefixes, revealing the mechanism of the impact of distributional discrepancy to the exploration dynamics in RLVR training and efficiently mitigates entropy collapse without requiring additional SFT, intricate reward designs, or complex prompting.
Xin Shen, Huishuai Zhang, Peng Li et al.· 0 citations
This work analyzes the decoding trajectories of LLaDA 2.0 and identifies a recurring diffusion confidence trap, which improves LLaDA 2.0 over confidence-based decoding, leading to more reliable mathematical reasoning.
Zhenhong Sun, Han-Qing Zhao, Yatao Bian et al.· 0 citations
The nature of test-time exploration in RLVR-trained LLMs is investigated by employing controlled maze-solving experiments and extracting a tree structure from mathematical reasoning traces (BODHI-Trees) based on semantic equivalence to delineate between entropy arising from stylistic variations and genuine inferential branching.
Soumadeep Saha, Krish Sharma, Akshay Chaturvedi et al.· 1 citation
Asynchronous reinforcement learning (RL) accelerates large language model (LLM) post-training by overlapping rollout generation with policy optimization, but the resulting stale, off-policy data can destabilize optimization and ultimately cause policy collapse. Existing methods typically retain or discard tokens based solely on the magnitude of their importance ratios, applying the same threshold uniformly across token positions. In this work, we reveal that the natural scale of the importance ratio varies systematically with token entropy. Under asynchronous dynamics, this entropy-ratio scaling dictates two distinct phenomena: at low entropy, the inherent train-inference discrepancy is drastically amplified into substantial sampling noise; at high entropy, in-flight weight updates naturally induce pronounced, legitimate exploratory deviations. Consequently, magnitude-only correction inadvertently admits the amplified noise while strictly masking out the essential exploration triggered by in-flight updates. To address this, we propose the Entropy-Scaled Trust Region (ESTR), which scales each token's off-policy deviation by its local entropy, requiring no auxiliary forward passes or explicit version-switch detection. Across long-horizon agentic tasks and mathematical reasoning benchmarks, ESTR consistently outperforms existing asynchronous methods and achieves the best train-inference consistency. It reaches $37.34$ avg@1 on BrowseComp-Plus and $95.69$ on multi-turn GSM8K, matching synchronous GRPO while achieving a $2.6\times$ speedup.
While Reinforcement Learning (RL) effectively incentivizes reasoning in Large Language Models, current pipelines are hindered by training instability and rapid entropy collapse. These limitations often stem from"Rollout Silencing"and low-quality gradient signals in standard sampling procedures. In this work, we propose a robust, data-centric framework to stabilize RL training. We first introduce Potential-Aware Query Mining (PAQM), which filters data dynamically to focus on the"Distillation Zone"---samples with high potential for capability elicitation. Furthermore, we present Hybrid Stratified Replay (HSR), a novel mechanism that restructures batches by stratifying rollouts based on Path Entropy, a rollout-level confidence proxy, and outcome reward. Within each optimization step, HSR reuses current-policy"Stability Anchors"and"Hard Negatives"to construct high-contrast optimization groups, then clears its buffers before the next step. This approach mitigates entropy collapse while improving the utilization of learning signals under limited compute. Our method outperforms strong baselines on complex reasoning tasks, offering a principled solution for stable and efficient RL fine-tuning.
Yi-Meng Ye, Shuang Chen, Wen-Xuan Huang et al.· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
