Shaolun Zhang

I'm a fourth-year undergraduate student at UC Berkeley. I'm involved in research at Berkeley Artificial Intelligence Research (BAIR).

Research

I'm broadly interested in deep learning, neural networks and its applications. In the past, I have worked on projects involving LLMs and multi-modal LLMs.

Interactive multi-round reasoning segmentation model that enhances LLM-based segmentation by exploiting conversational memory of both visual and textual outputs.

Theoretical analysis and empirical results demonstrating asymmetry in model weights leading to the Reversal Curse, where LLMs trained on "A is B" fails to learn "B is A".