cs.LGOct 9, 2026

Structure-Aware Numerical Stabilization for Half-Precision Proximal Policy Optimization

Authors: Linxuan Wu, Liang Chen

Organizations: Institute of Automation, Chinese Academy of Sciences School of Artificial Intelligence, University of Chinese Academy of Sciences Beijing, China

Read PDF
Ask AI
Cite

Copy citation

DOINot available for this paper

Abstract

Local online training on resource-constrained platforms requires low computational and training-state overhead. We study the numerical stability and computational efficiency of half-precision proximal policy optimization (PPO) on commodity GPUs. Directly converting PPO training to FP16 is unreliable because numerical errors feed back through policy updates and subsequent data collection, while PPO repeatedly compares a current policy with a fixed behavior policy. We propose PPO-FP16 with three structure-aware modules. Global scalar protection stabilizes statistics shared across a minibatch or parameter update. Policy-change residual protection extracts and preserves small differences between the current and behavior policies. Efficient persistent-update protection stabilizes optimizer state and parameter writeback while reducing state storage. These changes preserve the PPO objective, hyperparameters, and training budget. Across six continuous-control tasks and five seeds, PPO-FP16 achieves the lowest mean return-AUC rank (1.90) among the four methods and ranks first in 17 of 30 environment-seed comparisons. Eight of nine performance configurations yield PPO update speedups ranging from 1.16 to 2.03 times. Relative to Full FP32 Adam, optimizer-state memory is reduced by 71.7 to 74.9 percent, and peak allocated GPU memory is reduced by up to 44.5 percent. The results show that protecting a small set of structurally important quantities can deliver stable, efficient FP16 PPO without retuning.

Figures & tables

Figure 1. PPO-FP16 training pipeline and its three structure-aware protection modules. Module A protects global statistics and gradient scale, Module B protects small policy changes in the residual domain, and Module C stabilizes optimizer state. PPO-FP16 training pipeline and the three protection modules for global scalars, policy-change residuals, and persistent parameter updates.
Figure 2
More in this topic

RL Stability

See all 29 papers

Explore similar work

CardsList
Oct 1, 2026

Reusing Past Samples in Proximal Policy Optimization: When and How Does It Help?

Among on-policy deep reinforcement learning methods, Proximal Policy Optimization (PPO) has become the de facto standard, due to its consistently strong empirical performance across diverse application domains. However, on-policy methods are inherently sample inefficient: fresh data collected under the current policy is used for just a few updates before being discarded. Off-policy methods avoid this inefficiency via experience replay, achieving notable sample efficiency gains, but at the cost of training instabilities or extensive tuning. This motivated the rise of hybrid strategies that augment PPO with off-policy data reuse. Existing sample-reuse variants of PPO demonstrated improved sample efficiency over vanilla PPO, yet a systematic study of when reuse helps, in which scenarios, and to what extent remains missing. In this work, we study the effectiveness of sample reuse in PPO by instantiating two variants within a multiple importance weighting framework. Both retain the core PPO mechanics, reusing only samples from a window of recent iterations, thereby isolating the effect of data reuse from other factors. The variants, termed wPPO-U and wPPO-BH, employ vanilla importance weights or balance-heuristic-corrected ones, respectively. For both, we derive policy improvement lower bounds providing theoretical grounding for their respective losses. We use them to empirically study when and how data reuse improves sample efficiency or final performance of PPO across continuous control tasks.

Alessandro Montenegro, Riccardo Venturelli, Marco Mussi +2

Politecnico di Milano, Milan, Italy · Università degli Studi di Milano, Milan, Italy

Experience ReplayProximal Policy OptimizationPolicy LearningSample-Efficient RLOff-Policy RLImportance Sampling for RL
AskPDF
Sep 29, 2026

EasyPPO: Stabilizing the Critic Is Key

A key strength of Proximal Policy Optimization (PPO) is its learned critic, which uses historical trajectories collected during reinforcement learning to estimate expected returns and reduce policy-gradient variance. However, we find that the critic is also a major source of instability in reinforcement learning for large language models (LLMs). We identify two critic failure modes that destabilize PPO. First, filtering truncated rollouts from both actor and critic shifts the policy objective to reward conditioned on completion, allowing truncation to increase even as conditional reward improves. Second, heterogeneous return noise can cause high-variance prompts to dominate critic updates in finite batches. We introduce EasyPPO to address these failures. Actor-only overlong filtering trains the critic on returns from both completed and truncated rollouts. Noise-normalized critic regression weights each prompt's critic loss by the inverse standard deviation of its sampled returns, balancing noise contributions across prompts. Moderately smaller critic mini-batches confine outlier influence to fewer rollouts during gradient clipping. Across continuous-reward coding on FrontierCS, binary-reward mathematical reasoning on AIME24, and multi-turn search on Search-R1, EasyPPO remains stable throughout the full training horizon and consistently outperforms vanilla PPO, VAPO, and HL-Gauss PPO. Its best validation scores show relative gains of 14.89%, 2.28%, and 9.47% over PPO, respectively.

Xuanyi Zhou, Qiuyang Mang, Huanzhi Mao +7

University of California, Berkeley · Princeton University

RL for Language ModelsProximal Policy OptimizationActor-Critic MethodsRL for Language Model Reasoning
AskPDF
Mar 13, 2026

Optimize Wider, Not Deeper: Consensus Aggregation for Policy Optimization

Proximal policy optimization (PPO) approximates the trust region update using multiple epochs of clipped SGD. Each epoch may drift further from the natural gradient direction, creating path-dependent noise. To understand this drift, we can use Fisher information geometry to decompose policy updates into signal (the natural gradient projection) and waste (the Fisher-orthogonal residual that consumes trust region budget without first-order surrogate improvement). Empirically, signal saturates but waste grows with additional epochs, creating an optimization-depth dilemma. We propose Consensus Aggregation for Policy Optimization (CAPO), which redirects compute from depth to width: KK PPO replicates are optimized on the same batch, differing only in minibatch shuffling order, and then aggregated into a consensus. We study aggregation in two spaces: Euclidean parameter space, and the natural parameter space of the policy distribution via the logarithmic opinion pool. In natural parameter space, the consensus provably achieves higher KL-penalized surrogate and tighter trust region compliance than the mean expert; parameter averaging inherits these guarantees approximately. On continuous control tasks, CAPO outperforms PPO and compute-matched deeper baselines under fixed sample budgets by up to 8.6x. CAPO demonstrates that policy optimization can be improved by optimizing wider, rather than deeper, without additional environment interactions.

Zelal Su, Mustafaoglu, Sungyoung Lee +3

Lain · Department of Computer Science, University of Texas at Austin · Department of Electrical and Computer Engineering, University of Texas at Austin

Policy OptimizationProximal Policy OptimizationTrust Region Policy Optimization
AskPDF