My work follows two connected research directions: trustworthy learning over structured data, and efficient multimodal foundation models for video understanding.
A theoretical and empirical treatment of structural distribution shift in GNN explanations, using proxy graphs to preserve explanatory information while remaining in distribution.
Confidence-calibrated framework for reliable GNN explanations, especially under OOD settings.
Graph-enhanced RAG pipeline for evidence-grounded LLM responses.
A 3D GNN explainer that localizes explanations through node-wise radius of influence.
An explanation framework for graph regression models with improved reliability under distribution shift.
Uses LLM-guided Bayesian inference to mitigate learning bias in graph explanation.
Generates in-distribution proxy graphs to improve faithfulness of GNN explanations.
A data-augmentation-based framework to improve OOD robustness of GNN explanations.
Neural recovery of variable names and type inference from minified code.
Commit-level vulnerability detection and CVSS assessment via context-aware graph learning.
InfoShot is a training-free, shot-aware sampler that preserves both representative context and brief decision-critical events in long videos.
A two-stage self-improvement framework that uses model-generated rationales to adapt vision-language models to domain-specific video classification without additional annotations.