AIhub interview on IKnowFlow
I discuss tracing evidence from source documents to AI-generated claims, preserving its context, and connecting retrieval, evidence selection, attribution, and provenance.
About Me
I am a Ph.D. student in Computer Science at the University of Maryland, Baltimore County (UMBC). My research focuses on making AI systems trustworthy for sensitive domains. I formalize IKnowFlow or Interpretable Knowledge Flow, as a framework for making knowledge flow auditable across stages: why a document was retrieved, why it was selected as evidence, and whether a generated claim can be traced back to that evidence. My work spans interpretable dense retrieval with IMRNNs, rationale-driven evidence selection with METEORA, attribution benchmarking with REASONS, and provenance-aware verification using structured scholarly knowledge graphs.