English static mirror for SEO/GEO · AI-assisted translation · Read Chinese original

Score Centering Stabilizes Off-policy Reinforcement Learning

Forum topic · 小凯 · 2026-09-19

Summary

This paper (arXiv:2609.20807) by Martin Marek and Max Ryabinin addresses training-inference mismatch (TIM) in reinforcement learning for large language models. The authors show that RL instability under TIM is primarily caused by drift: a persistent bias between training and inference engines that accumulates with each training step. They derive an additive 'score centering' correction term that stabilizes RL under TIM by canceling drift. In experiments training models from 0.6B to 30B parameters, score centering alone matches or outperforms importance sampling methods based on quantization, with the performance gap growing as mismatch becomes more severe. Because the correction is additive, score centering can be combined with importance sampling; in staleness experiments, the combination outperforms pure importance sampling baselines. The approach offers a practical alternative to fully eliminating TIM, which would come at a major cost to rollout efficiency.

Paper Overview

Field: Machine Learning Authors: Martin Marek, Max Ryabinin Published: 2026-09-17 arXiv: 2609.20807

Abstract

Reinforcement learning (RL) of large language models is notoriously sensitive to small differences between training and inference engines, often referred to as the training-inference mismatch (TIM). However, completely eliminating TIM is impractical, as it would come at a major cost to rollout efficiency. In this paper, we show that the instability of RL under TIM is primarily caused by drift: a persistent bias between training and inference engines that accumulates with every training step. We derive an additive "score centering" correction term that stabilizes RL under TIM by canceling drift. When training models from 0.6B to 30B parameters, score centering alone matches or outperforms methods based on importance sampling under quantization, with the gap growing as the mismatch becomes more severe. Since the correction is additive, score centering can also be combined with importance sampling — in our staleness experiments, their combination outperforms pure importance sampling baselines.

Key Contributions

  • Identifies drift (a persistent, accumulating bias between training and inference engines) as the primary cause of RL instability under training-inference mismatch (TIM).
  • Derives an additive score centering correction term that cancels drift and stabilizes off-policy RL.
  • Demonstrates at scale (0.6B–30B parameter models) that score centering alone matches or beats importance sampling based on quantization, with the advantage growing as mismatch increases.
  • Shows that score centering composes with importance sampling, outperforming pure importance sampling baselines in staleness experiments.
--- *Auto-collected on 2026-09-19*

Tags

#reinforcement-learning#large-language-models#training-inference-mismatch#score-centering#importance-sampling#off-policy-rl#arxiv

This page is an English static mirror generated for search and AI citation. It may be a full translation or structured summary of the Chinese original. Canonical interactive discussion lives on the Chinese page: https://zhichai.net/topic/178634982