Shiliang Wu

Identifiers

• name variant Shiliang Wu 0.60 · backfill

Papers (1)

1. Decomposing MXFP4 quantization error for LLM reinforcement learning: reducible bias, recoverable deadzone, and an irreducible floor cs.LG · 2026 · author #2

Mentions

• 2605.20402 #2 · arxiv_oai · confidence 0.70 Shiliang Wu

Frequent Coauthors

• Xiaocan Li 1 shared papers
• Zheng Shen 1 shared papers