17 papers
Distilled Reinforcement Learning for LLM Post-training
Chen Wang, Zhaochun Li, Jionghao Bai +4
Large language model (LLM) post-training is essential for improving reasoning, adaptation, and alignment. Existing methods mainly follow two paradigms: reinforcement learning (RL)…
SCOPE-RL: Stable and Quantitative Control of Policy Entropy in RL Post-Training
Chen Wang, Zhaochun Li, Jionghao Bai +3
Reinforcement learning (RL) is a key paradigm for post-training large language models (LLMs), but the widely used Group Relative Policy Optimization (GRPO) often suffers from entro…
Implicit Compression Regularization: Concise Reasoning via Internal Shorter Distributions in RL Post-Training
Chen Wang, Hexuan Deng, Yining Zhang +5
Reinforcement learning with verifiable rewards improves LLM reasoning but often induces overthinking, where models generate unnecessarily long reasoning traces. Existing methods ma…
HieraSparse: Hierarchical Semi-Structured Sparse KV Attention
Haoxuan Wang, Chen Wang
The deployment of long-context Large Language Models (LLMs) poses significant challenges due to the intense computational cost of self-attention and the substantial memory overhead…
Targeted Exploration via Unified Entropy Control for Reinforcement Learning
Chen Wang, Lai Wei, Yanzhi Zhang +5
Recent advances in reinforcement learning (RL) have improved the reasoning capabilities of large language models (LLMs) and vision-language models (VLMs). However, the widely used…
PipeLive: Efficient Live In-place Pipeline Parallelism Reconfiguration for Dynamic LLM Serving
Xu Bai, Muhammed Tawfiqul Islam, Chen Wang +1
Pipeline parallelism (PP) is widely used to partition layers of large language models (LLMs) across GPUs, enabling scalable inference for large models. However, existing systems re…