QaRL: Rollout-Aligned Quantization-Aware RL for Fast and Stable Training under Training--Inference Mismatch
2026-04-09 • Machine Learning
Machine LearningArtificial Intelligence
AI summaryⓘ
The authors address a slowdown in training large language models when using reinforcement learning because generating example outputs (rollouts) takes a long time. To speed this up, others have used low-precision math during rollouts, but this causes issues since training updates use high precision, creating a mismatch. The authors propose QaRL, which aligns training computations with the low-precision rollouts to reduce this mismatch. They also identify errors in long generated responses and introduce TBPO, a method to keep training updates stable. Their approach improves performance and stability while keeping the speed benefits of low-precision rollouts.
large language modelsreinforcement learningrolloutsquantizationtraining-inference gapprecisionpolicy optimizationtrust regionMoE modelssequence-level objectives
Authors
Hao Gu, Hao Wang, Jiacheng Liu, Lujun Li, Qiyuan Zhu, Bei Liu, Binxing Xu, Lei Wang, Xintong Yang, Sida Lin, Sirui Han, Yike Guo
Abstract
Large language model (LLM) reinforcement learning (RL) pipelines are often bottlenecked by rollout generation, making end-to-end training slow. Recent work mitigates this by running rollouts with quantization to accelerate decoding, which is the most expensive stage of the RL loop. However, these setups destabilize optimization by amplifying the training-inference gap: rollouts are operated at low precision, while learning updates are computed at full precision. To address this challenge, we propose QaRL (Rollout Alignment Quantization-Aware RL), which aligns training-side forward with the quantized rollout to minimize mismatch. We further identify a failure mode in quantized rollouts: long-form responses tend to produce repetitive, garbled tokens (error tokens). To mitigate these problems, we introduce TBPO (Trust-Band Policy Optimization), a sequence-level objective with dual clipping for negative samples, aimed at keeping updates within the trust region. On Qwen3-30B-A3B MoE for math problems, QaRL outperforms quantized-rollout training by +5.5 while improving stability and preserving low-bit throughput benefits.