FourSight: Stable FP4 RL Post-Training via Sequence Filtering and Sparse Channel Compensation
Reinforcement learning (RL) post-training increasingly targets low-bit deployment, with NVFP4 weight-and activation quantization (W4A4) becoming the state-of-the-practice format. However, integrating NVFP4 into RL degrades rollout quality and amplifies cross-engine training-inference mismatch (TIM), destabilizing train...
Yi-Chen Wei, Tian-Yuan Wu, Gu Gong et al.
· 0 citations