Liqing Zhang

Even the best LLMs struggle to navigate the fine line between empathetic support and harmful validation in emotionally charged Bengali conversations.

Reconstruct a high-fidelity, full-head 3D avatar from a single image in under one second, finally breaking the quality-speed tradeoff.

MLLMs can achieve state-of-the-art multimodal retrieval by learning to compress information into a handful of "bottleneck" tokens, forcing the model to distill relevant semantics.