Zhao Yang (杨钊)

I am a fourth-year PhD student at the Gaoling School of Artificial Intelligence, Renmin University of China, supervised by Prof. Bing Su. I also work closely with Prof. Chuan Cao.

My research interests focus on Large Language Models and Diffusion Models for modeling sequence data, especially biological languages.

yangyz1230@gmail.com  /  Google Scholar  /  GitHub

Research

(* indicates equal contribution)

2026

Through experiments, we find that current gene expression predictors have limited long-sequence modeling ability; Prism effectively integrates multimodal epigenomic signals with DNA sequences to achieve state-of-the-art performance.

We design a unique interface that enables LLMs to directly understand DNA sequences, then introduce CRE-ReasonBench and a biological reasoning-informed regression framework for interpretable regulatory DNA activity prediction.

2025

We show that sequence-to-function genomic profile predictors can learn strong and transferable DNA representations, then leverage MoE to better model cross-species and cross-profile information.

We propose a reinforcement learning framework for designing regulatory DNA sequences using transcription factor binding site rewards.

2023

A text-driven diffusion-based framework for synthesizing long-term human motions, achieving coherent and realistic sequences.

We present D3LM, a discrete DNA diffusion language model that unifies bidirectional DNA understanding and DNA generation.

We investigate the ability of diffusion-based language models to implicitly determine the optimal infilling lengths.

A sequence-based science foundation model that unifies molecules, proteins, DNA and RNA for cross-domain scientific discovery.

A hybrid Transformer-Mamba2 long-context DNA language model for genomic understanding and generation.

We propose ProtDETR, a novel framework that reframes enzyme function prediction as an object detection problem by identifying active sites as "objects".

We propose MoMu, a foundation model associating molecular graphs with natural language, enabling multimodal understanding in molecular science.

Template from Jon Barron.