Pratibha Zunjare

Hi! I am a PhD student at Virginia Tech. My research focuses on large language models (LLMs) and their capacity for structured reasoning. Currently, I am working on multi document data reasoning. In parallel, I am analyzing the limitations of LLMs in following structured reasoning steps within RAG-based workflows, aiming to understand and address the root causes behind their failures on tasks requiring systematic thought. Currently, I am focusing on exiting directions:

• Evaluating LLMs on multi-document reasoning tasks and dataset development

• Subspace embedding–based key-value cache compression for optimizing LLM performance

• Neuro-symbolic reasoning for healthcare applications

Previously, I completed my Master’s degree in Computer Science at the National Institute of Technology Rourkela. I also bring over seven years of software engineering experience across a range of domains. My broader research goal is to deepen my understanding of how language models process, interpret, and reason through natural language. My research is centered around several key areas:

• Interpretability of Large Language Models: Gaining insights into the internal mechanisms and reasoning processes of LLMs.

• Agentic Frameworks and Tool Use: Designing efficient agent-based workflows that enable collaborative reasoning and intelligent tool utilization in LLMs.

• Evaluation and Benchmarking: Creating innovative and rigorous evaluation methods to assess LLM performance in complex and challenging scenarios.

• Knowledge Representation in LLMs: Developing novel strategies to mitigate hallucinations by improving how LLMs represent and integrate knowledge.

If you’re interested in my research, would like to discuss related topics, or explore potential collaborations, I’d be happy to connect. The best way to reach me is via email at pratibhaz@vt.edu

📰 News