一作ACL 2026Poster· Natural Language Processing
Sparse- RL : Breaking the Memory Wall in LLM Reinforcement Learning via Stable Sparse Rollouts
Sijia Luo · Xiaokang Zhang · Yuxuan Hu · Bohan Zhang · Ke Wang · Jinbo Su · Mengshu Sun · Lei Liang · Jing Zhang
摘要
Reinforcement Learning (RL) has become essential for eliciting complex reasoning capabilities in Large Language Models (LLMs). However, the substantial memory overhead of storing Key-Value (KV) caches during long-horizon rollouts acts as a critical bottleneck, often prohibiting efficient training...