5 papers
Beyond Euclidean Clipping: Overcoming Exploration Collapse in LLM RL via Riemannian Isometric Policy Optimization
Zhicheng Cai, Xinyuan Guo, Hanlin Wu +4
Reinforcement learning (RL) has become a dominant paradigm for enhancing LLMs' reasoning capabilities. However, RL algorithms with PPO-Clip are inherently limited by exploration co…
Weak-to-Strong Generalization via Direct On-Policy Distillation
Shiyuan Feng, Huan-ang Gao, Haohan Chi +7
Reinforcement learning with verifiable rewards (RLVR) is a powerful recipe for improving language-model reasoning, but it is expensive to repeat on every new strong model because t…
Spectral Rewiring for Exploration, Purification, and Model Merging
Zhilong Zhang, Hongli Yu, Huan-ang Gao +5
Reinforcement learning has become a standard post-training recipe for large language models, but dense full-parameter updates create two deployment-relevant bottlenecks: suppressed…
CUDA Agent: Large-Scale Agentic RL for High-Performance CUDA Kernel Generation
Weinan Dai, Hanlin Wu, Qiying Yu +13
GPU kernel optimization is fundamental to modern deep learning but remains a highly specialized task requiring deep hardware expertise. Despite strong performance in general progra…
RouteMoA: Dynamic Routing without Pre-Inference Boosts Efficient Mixture-of-Agents
Jize Wang, Han Wu, Zhiyuan You +9
Mixture-of-Agents (MoA) improves LLM performance through layered collaboration, but its dense topology raises costs and latency. Existing methods employ LLM judges to filter respon…