Selected Publications
Contribution: Proposed a causal intervention framework to mitigate logical hallucinations in LLMs.
Contribution: Developed a multimodal RAG system for generating accurate radiology reports.
Contribution: Created the first comprehensive benchmark for anomaly detection in NLP tasks.
Contribution: Evaluated LLMs' capabilities in detecting anomalies across various domains.
Open Source & Engineering
A framework for evaluating and optimizing agents and LLMs. Supports parallel experiments in thousands of environments and RL rollout generation.
A distributed machine learning platform for scalable model training and hyperparameter tuning. Optimized resource scheduling and execution efficiency.