# When Importance Sampling Misallocates Credit: Asymmetric Ratios for Outcome-Supervised RL

arXiv: https://arxiv.org/abs/2510.06062 · HTML: https://mlsanity.com/paper/2510.06062/

## Key facts

- Cited by 21 papers in the map (data through 14 Oct 2026).
- Builds on 29 of its 39 references that are in the map.
- First posted on arXiv 7 Oct 2025.

## Builds on (most cited first, 29 of 29)

- [Training language models to follow instructions with human feedback](https://mlsanity.com/paper/2203.02155/) (2022)
- [Proximal Policy Optimization Algorithms](https://mlsanity.com/paper/1707.06347/) (2017)
- [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://mlsanity.com/paper/2402.03300/) (2024)
- [Qwen3 Technical Report](https://mlsanity.com/paper/2505.09388/) (2025)
- [DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning](https://mlsanity.com/paper/2501.12948/) (2025)
- [Efficient Memory Management for Large Language Model Serving with PagedAttention](https://mlsanity.com/paper/2309.06180/) (2023)
- [Deep reinforcement learning from human preferences](https://mlsanity.com/paper/1706.03741/) (2017)
- [Let's Verify Step by Step](https://mlsanity.com/paper/2305.20050/) (2023)
- [Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback](https://mlsanity.com/paper/2204.05862/) (2022)
- [DAPO: An Open-Source LLM Reinforcement Learning System at Scale](https://mlsanity.com/paper/2503.14476/) (2025)
- [HybridFlow: A Flexible and Efficient RLHF Framework](https://mlsanity.com/paper/2409.19256/) (2024)
- [LiveCodeBench: Holistic and Contamination Free Evaluation of Large Language Models for Code](https://mlsanity.com/paper/2403.07974/) (2024)
- [Competition-Level Code Generation with AlphaCode](https://mlsanity.com/paper/2203.07814/) (2022)
- [Solving Quantitative Reasoning Problems with Language Models](https://mlsanity.com/paper/2206.14858/) (2022)
- [OlympiadBench: A Challenging Benchmark for Promoting AGI with Olympiad-Level Bilingual Multimodal Scientific Problems](https://mlsanity.com/paper/2402.14008/) (2024)
- [Group Sequence Policy Optimization](https://mlsanity.com/paper/2507.18071/) (2025)
- [VAPO: Efficient and Reliable Reinforcement Learning for Advanced Reasoning Tasks](https://mlsanity.com/paper/2504.05118/) (2025)
- [MiniMax-M1: Scaling Test-Time Compute Efficiently with Lightning Attention](https://mlsanity.com/paper/2506.13585/) (2025)
- [ProRL: Prolonged Reinforcement Learning Expands Reasoning Boundaries in Large Language Models](https://mlsanity.com/paper/2505.24864/) (2025)
- [A Survey of Reinforcement Learning for Large Reasoning Models](https://mlsanity.com/paper/2509.08827/) (2025)
- [Skywork Open Reasoner 1 Technical Report](https://mlsanity.com/paper/2505.22312/) (2025)
- [Mastering Complex Control in MOBA Games with Deep Reinforcement Learning](https://mlsanity.com/paper/1912.09729/) (2019)
- [Towards Human-Level Bimanual Dexterous Manipulation with Reinforcement Learning](https://mlsanity.com/paper/2206.08686/) (2022)
- [DCPO: Dynamic Clipping Policy Optimization](https://mlsanity.com/paper/2509.02333/) (2025)
- [FlowRL: Matching Reward Distributions for LLM Reasoning](https://mlsanity.com/paper/2509.15207/) (2025)
- [Stabilizing Knowledge, Promoting Reasoning: Dual-Token Constraints for RLVR](https://mlsanity.com/paper/2507.15778/) (2025)
- [Klear-Reasoner: Advancing Reasoning Capability via Gradient-Preserving Clipping Policy Optimization](https://mlsanity.com/paper/2508.07629/) (2025)
- [CE-GPPO: Coordinating Entropy via Gradient-Preserving Clipping Policy Optimization in Reinforcement Learning](https://mlsanity.com/paper/2509.20712/) (2025)
- [Attention as a Compass: Efficient Exploration for Process-Supervised RL in Reasoning Models](https://mlsanity.com/paper/2509.26628/) (2025)

## Built on by (most cited first, 21 of 21)

- [MASPO: Unifying Gradient Utilization, Probability Mass, and Signal Reliability for Robust and Sample-Efficient LLM Reasoning](https://mlsanity.com/paper/2602.17550/) (2026)
- [When Right Meets Wrong: Bilateral Context Conditioning with Reward-Confidence Correction for GRPO](https://mlsanity.com/paper/2603.13134/) (2026)
- [STAPO: Stabilizing Reinforcement Learning for LLMs by Silencing Rare Spurious Tokens](https://mlsanity.com/paper/2602.15620/) (2026)
- [Accelerating RL Post-Training Rollouts via System-Integrated Speculative Decoding](https://mlsanity.com/paper/2604.26779/) (2026)
- [Rethinking Muon Beyond Pretraining: Spectral Failures and High-Pass Remedies for VLA and RLVR](https://mlsanity.com/paper/2605.19282/) (2026)
- [A Step Back: Prefix Importance Ratio Stabilizes Policy Optimization](https://mlsanity.com/paper/2601.22718/) (2026)
- [QaRL: Rollout-Aligned Quantization-Aware RL for Fast and Stable Training under Training--Inference Mismatch](https://mlsanity.com/paper/2604.07853/) (2026)
- [Ratio-Variance Regularized Policy Optimization for Efficient LLM Fine-tuning](https://mlsanity.com/paper/2601.03320/) (2026)
- [From log π to π: Taming Divergence in Soft Clipping via Bilateral Decoupled Decay of Probability Gradient Weight](https://mlsanity.com/paper/2603.14389/) (2026)
- [Putting on the Thinking Hats: A Survey on Chain of Thought Fine-tuning from the Perspective of Human Reasoning Mechanism](https://mlsanity.com/paper/2510.13170/) (2025)
- [Asymmetric Prompt Weighting for Reinforcement Learning with Verifiable Rewards](https://mlsanity.com/paper/2602.11128/) (2026)
- [From AI Assistant to AI Scientist: Autonomous Discovery of LLM-RL Algorithms with LLM Agents](https://mlsanity.com/paper/2603.23951/) (2026)
- [Balanced Aggregation: Understanding and Fixing Aggregation Bias in GRPO](https://mlsanity.com/paper/2605.04077/) (2026)
- [The Extrapolation Cliff in On-Policy Distillation of Near-Deterministic Structured Outputs](https://mlsanity.com/paper/2605.08737/) (2026)
- [PCL-Reasoner-V1.5: Advancing Math Reasoning with Offline Reinforcement Learning](https://mlsanity.com/paper/2601.14716/) (2026)
- [Nonsense Helps: Prompt Space Perturbation Broadens Reasoning Exploration](https://mlsanity.com/paper/2605.05566/) (2026)
- [Smooth Gate Functions for Soft Advantage Policy Optimization](https://mlsanity.com/paper/2602.19345/) (2026)
- [MHPO: Modulated Hazard-aware Policy Optimization for Stable Reinforcement Learning](https://mlsanity.com/paper/2603.16929/) (2026)
- [OGER: A Robust Offline-Guided Exploration Reward for Hybrid Reinforcement Learning](https://mlsanity.com/paper/2604.18530/) (2026)
- [Bounded Ratio Reinforcement Learning](https://mlsanity.com/paper/2604.18578/) (2026)
- [ReCo: Reweighting GRPO Against Distributional Concentration](https://mlsanity.com/paper/2607.26862/) (2026)

## Abstract (arXiv)

Reinforcement learning (RL) has shown great promise in large language models (LLMs) post-training, which typically rely on token-level clipping to maintain stability during optimization. Despite the empirical success of GRPO-style methods, we identify a fundamental and previously overlooked challenge in this popular Outcome-Supervised RL (OSRL) paradigm. We reveal that in OSRL, where advantages are shared across tokens within a response, importance sampling (IS) ratios deviate from their traditional purpose of distribution correction as in classic RL, which become token-level weights that allocate the shared advantage signal across tokens. We show that this hidden role shift induces a critical mismatch for positive-advantage tokens, leading to unbalanced token weighting between positive and negative tokens. Specifically, it suppresses the update of underrepresented tokens that are lagging behind, while over-amplifying already high-probability tokens. This mismatch results in rich-get-richer dynamics that over-reinforce confident tokens, weaken catch-up learning that drive entropy collapse, excessive repetition, and premature convergence. To address this, we propose Asymmetric Importance Sampling Policy Optimization (ASPO), a simple yet effective strategy that reverses the ratio-induced weighting of positive-advantage tokens, while stabilizing extreme updates and maintaining gradient flow. This mismatch correction aligns their update direction with the learning dynamics of negative ones. Comprehensive experiments across math reasoning and coding benchmarks demonstrate that ASPO significantly mitigates entropy collapse, improves training stability, and enhances performance over strong GRPO-based baselines. Our analysis provides new insights into the role of token-level weighting in OSRL and highlights the critical importance of correcting ratio-induced weighting in LLM RL.