Composition‑RL is an offline data‑augmentation method that composes existing prompts before reinforcement learning to address the pass‑rate‑1 bottleneck in large‑language‑model training. By generating and verifying a composed dataset once from a fixed source pool, it avoids modifying the RL optimizer and improves reasoning performance across models ranging from 4B to 30B. A curriculum variant that gradually increases compositional depth further boosts performance, and the approach also facilitates cross‑domain RL by combining prompts from different domains.
By Xin Xu, Clive Bai, Kai Yang, Tianhao Chen, Cai Yuchen, Yangkun Chen, Weijie Liu, Hao Chen, Yang Wang, Saiyong Yang, Can Yang
The paper "Demystifying Reinforcement Learning Post-Training of Language Models" investigates how reinforcement learning (RL) post‑training enhances large language models (LLMs) for tasks such as reasoning, math, and coding. By isolating RL components in a controlled setting, the authors analyze how the base model’s prior distribution, reward granularity, prompt diversity, and model scale influence outcomes, using policy entropy to compare pre‑training, supervised fine‑tuning (SFT), and RL stages. The study clarifies the role of spurious rewards, the importance of the base model’s probability mass on desired behaviors, and how these factors interact to determine post‑training success, offering a practical primer for NLP researchers.
"whyItMatters":"The work provides a clearer understanding of RL post‑training mechanics, helping researchers and practitioners effectively apply RL to improve LLM capabilities."
By Donovan Clay, Saket Gollapudi, Sankar Harilal, Min Jang, Jacob Morrison, Sewoong Oh, Natasha Jaques
arXiv:2606. 04516v1 Announce Type: cross Abstract: Reinforcement learning with verifiable rewards (RLVR) significantly advances LLM reasoning, yet it faces a dilemma: standard supervised scaling is throttled by high annotation costs, while unsupervised alternatives suffer from severe model collapse.
By Guangcheng Zhu, Shenzhi Yang, Haobo Wang, Xing Zheng, Yingfan MA, Xuening Feng, Zhongqi Chen, Kai Tang, Zhengqing Zang, Bowen Song, Weiqiang Wang, Gang Chen
arXiv:2603. 16140v2 Announce Type: replace Abstract: Reinforcement learning with verifiable rewards (RLVR) has driven recent capability advances of large language models across various domains.
By Yuxuan Zhu, Daniel Kang
arXiv:2610.02015v1 Announce Type: cross
Abstract: Recent advances in LLM reasoning models---driven primarily by the paradigm of post-training via reinforcement learning with verifiable reward (RLVR)-...
By Michael Sullivan, Alexander Koller
arXiv:2609.24243v3 Announce Type: replace
Abstract: Reinforcement learning (RL) has substantially improved the reasoning capabilities of vision-language models (VLMs), yet it can induce chain-of-thou...
By Xutao Mao, Jianing Zhu, Jinman Zhao, Tongliang Liu, Xiaowen Chu, Cong Wang, Bo Han
arXiv:2607. 01181v1 Announce Type: cross Abstract: RL with verifiable rewards (RLVR) has emerged as a powerful paradigm for training LMs on tasks with well-defined success metrics, such as code generation and mathematical reasoning.
By Mehul Damani, Isha Puri, Idan Shenfeld, Jacob Andreas
arXiv:2605. 21125v2 Announce Type: replace Abstract: Group Relative Policy Optimization (GRPO), a prominent algorithm within the Reinforcement Learning from Verifiable Rewards (RLVR) framework, has achieved strong results in improving the reasoning capabilities of large language models (LLMs).
By Xixiang He, Qiyao Sun, Ao Cheng, Xingming Li, Xuanyu Ji, Hailun Lu, Runke Huang, Qingyong Hu
The paper introduces Calibrated Clipping, a dynamic method to align FP8 quantization bounds with high‑precision BF16 distributions, thereby mitigating training instability in full‑pipeline FP8 reinforcement learning for large language models. It identifies that compounded FP8 noise distorts importance ratios, causing entropy surges and garbled outputs. Experiments across GRPO and DAPO algorithms on 8B‑32B models show the technique restores performance to BF16 levels.
By Fanchao Chen, Ziheng Jiang, Ziyun Wei, Zheng Zhong, Du Li, Chi Zhang, Haibin Lin, Shivaram Venkataraman
arXiv:2610.22723v1 Announce Type: new
Abstract: Large language models exhibit misbehaviors that hinder deployment in high-stakes settings. One approach to monitor such failures is to use an external...
By Guy Bar-Shalom, Mor Geva, Fabrizio Frasca, Haggai Maron
DE‑Venus is a unified, data‑efficient framework for reinforcement learning with verifiable rewards (RLVR) tailored to large language models. It structures the RLVR lifecycle into three modules—Active Data Selection, Weak Supervision Construction, and Training‑Time Supervision Refinement—allowing method‑specific decisions to be expressed as dataset transitions or online transformations while maintaining distributed execution contracts. Experiments on public benchmarks and three business scenarios show that DE‑Venus can preserve or improve model quality using only 10% of labels or 13% of relevant data, and can cut convergence steps by 63%–75% in selected business configurations.
By Shenzhi Yang, Guangcheng Zhu, Kai Tang, Zhengqing Zang, Xing Zheng, Haobo Wang, Yingfan Ma, Bowen Song, Bo Han, Bo An, Lei Feng, Weiqiang Wang, Junbo Zhao, Gang Chen
arXiv:2608. 08802v1 Announce Type: new Abstract: Reinforcement Learning with Verifiable Rewards (RLVR) makes Multimodal Large Language Models more accurate, but the gains are brittle: simply paraphrasing a question or changing the prompt template can degrade them, which challenges reliable deployment in high-stakes scenarios like medical VQA.
By Pengfei Zhou, Zhiwei Tang, Xiaopeng Peng, Chenrui Zhou, Lama Moukheiber, Yixing Ma, Bin Xu, Jiajun Song, Zhenglin Wan, Wangbo Zhao, Jiasheng Tang, Bohan Zhuang, Fan Wang, Yang You