Pengxiang Cheng

Pengxiang Cheng

I am a research scientist and team lead in the AI group at Bloomberg, where I manage the Core NLP team, focusing on developing libraries and frameworks for NLP and LLM applications and training foundation models with financial domain knowledge.

Prior to joining Bloomberg, I obtained my Ph.D. in Computer Science at UT Austin, working on natural language understanding and computational semantics with Dr. Katrin Erk. I completed my undergraduate studies at  Tsinghua University, majoring in Automation and Economics.

Publications

Unsupervised Contrast-Consistent Ranking with Language Models.

Details

Overcoming Catastrophic Forgetting in Massively Multilingual Continual Learning.

Details

Dataless Knowledge Fusion by Merging Weights of Language Models.

Code Details