[2510.11696] QeRL: Beyond Efficiency -- Quantization-enhanced Reinforcement Learning for LLMs
Abstract:We propose QeRL, a Quantization-enhanced Reinforcement Learning framework for large language models (LLMs). While RL is essential for LLMs' reasoning capabilities, it is resource-intensive, requiring substantial GPU memory and long rollout durations. QeRL addresses these issues by combining NVFP4 quantization with Low-Rank Adaptation (LoRA), accelerating rollout phase of RL while reducing memory overhead. Beyond efficiency, our findings show that quantization noise increases policy entropy, enhancing exploration, and enabling the discovery of better strategies during RL. To further optimize exploration, QeRL introduces an Adaptive Quantization Noise (AQN) mechanism, which dynamically adjusts noise during training. Experiments demonstrate that QeRL delivers over 1.5 times speedup in the rollout phase. Moreover, this is the first framework to enable RL training of a 32B LLM on a single H100 80GB GPU, while delivering overall speedups for RL training. It also achieves faster reward growth and higher final accuracy than 16-bit LoRA and QLoRA, while matching the performance of full-parameter fine-tuning on mathematical benchmarks such as GSM8K (90.8%) and MATH 500 (77.4%) in the 7B model. These results establish QeRL as an efficient and effective framework for RL training in LLMs.
Q E RL:B EYOND E FFICIENCY – Q UANTIZATION - ENHANCED R EINFORCEMENT L EARNING FOR LLM S Wei Huang1,3 Yi Ge2,4 Shuai Yang1 Yicheng Xiao4 Huizi Mao1 Yujun Lin1 Hanrong Ye1 Sifei Liu1 Ka Chun Cheung1 Hongxu Yin1 Yao Lu1 Xiaojuan Qi3 Song Han1,2 Yukang Chen1 1 2 3 4…
saved by
related reading
- [2504.13837] Does Reinforcement Learning Really Incentivize Reasoning Capacity in LLMs Beyond the Base Model?arxiv.org
- 2305.14314arxiv.org
- LoRA Without Regret - Thinking Machines Labthinkingmachines.ai
- State of RL for reasoning LLMs | A. Weersaweers.de
- How can LLM RL Work Despite Information-Theoretic Inefficiencyberen.io
- A Guide to Quantization in LLMs | Symbl.aisymbl.ai
- LLM Resourcesforrestbicker.com
- The 4-bitter Lesson | humans&humansand.ai
- What Makes Low-Bit Quantization-Aware Training Work for Reasoning LLMs? A Systematic Studyarxiv.org
- Quantization from the ground upngrok.com
- Efficient LLM inferencefinbarrtimbers.substack.com
- RL Scaling Laws for LLMs - by Cameron R. Wolfe, Ph.D.cameronrwolfe.substack.com