Selected work
Research and engineering projects across multimodal learning, graph intelligence, computer vision, and language.
Inverse-LLaVA projects text embeddings into continuous visual space and fuses the modalities within intermediate transformer layers, enabling multimodal reasoning without explicit alignment pre-training.
Take a look
A fusion network that combines Graph Neural Network representations with LLMs through knowledge injection and PEFT to generate textual edge features in social networks.
Take a look
LLM systems that code negotiation transcripts and simulate realistic negotiation scenarios for research.
Take a look
A Vision Transformer pipeline for classifying ancient mortar particles and comparing material composition across archaeological samples.
Take a look
