Research

Large Language Models (LLMs)

• Explaining Large Language Models Decisions using Shapley Values (2024)
  • Developed a model-agnostic framework for interpreting LLMs’ decisions
  • Created a novel method for prompt optimization and prompt engineering
  • Discovered and characterized the token noise effect, useful for designing LLM-powered recommendation systems
  • Video Presentation
• Creativity Has Left the Chat: The Price of Debiasing Language Models (2024)
  • Demonstrated the trade-offs between safety and creativity of LLMs
  • Built semantic diversity measurement using SentenceBERT and t-SNE
  • Identified RLHF as the root cause of LLM mode collapse (lack of creativity)
  • Proposed architectural solutions to preserve model creativity post-alignment
    • Implemented proof-of-concept using Apple’s MLX framework (CUDA alternative for Apple Silicon)
  • Video Presentation

AI Economics & Regulation

• Regulating eXplainable AI (XAI) May Harm Consumers (INFORMS Marketing Science, 2022)
  • Conducted the first comprehensive analysis of XAI regulation impact on business operations and consumer welfare
  • Developed strategies for companies to maintain competitive advantage while complying with AI regulations
  • Created framework for businesses to optimize AI transparency based on market conditions
• Analyzing the Role of Apple-Samsung Lawsuits on Pricing and Marketing Strategies of the Two Companies (New Marketing Research Journal, 2019)