Zuxin Liu
Scholar

Zuxin Liu

Google Scholar ID: 5ApCTCoAAAAJ
Salesforce AI Research
LLM agentreinforcement learningoptimizationembodied intelligence
Citations & Impact
All-time
Citations
2,679
 
H-index
23
 
i10-index
32
 
Publications
20
 
Co-authors
33
list available
Resume (English only)
Academic Achievements
  • Led development of the xLAM family of Large Action Models (1B to 8x22B), achieving #1 rankings on the Berkeley Function-Calling Leaderboard
  • Proposed APIGen-MT, an agentic multi-turn data synthesis pipeline powering xLAM-2
  • Papers accepted at top venues including ICLR 2025, NeurIPS 2024, ICML 2024, AISTATS 2024, and DMLR Journal
  • Released AgentLite, a multi-LLM-agent framework
  • Built and open-sourced leading function-calling datasets and models; APIGen dataset ranked Top-3 trending on HuggingFace
  • Published multiple works on safe reinforcement learning, offline RL, robustness certification, and efficient foundation model adaptation