Guanghan Wang

prof_pic.jpg

I am a Computer Science Ph.D. candidate at Cornell University (Cornell Tech) in New York City, advised by Volodymyr Kuleshov.

My research focuses on developing principled and efficient training and inference algorithms for foundation models, with an emphasis on generative modeling. My work spans pretraining, reinforcement learning, and inference-time generation.

I am currently a Research Intern at Amazon, where I have been studying the pretraining efficiency and scaling laws of diffusion and autoregressive language models.

Previously, I received my B.Eng. in Computer Science and Technology from Tsinghua University in 2024 and was a visiting research student at Carnegie Mellon University in 2023.

news

Oct 13, 2026 Our new preprint “Revisiting the 16× Pretraining Compute Gap Between Diffusion and Autoregressive Language Models” is now available on arXiv.
Apr 30, 2026 d2 has been accepted to ICML 2026.
Feb 16, 2026 I joined Amazon as an Applied Scientist Intern, working on diffusion language model pretraining and scaling laws.
Sep 18, 2025 ReMDM has been accepted to NeurIPS 2025.
Jan 22, 2025 Our paper “Simple Guidance Mechanisms for Discrete Diffusion Models” has been accepted to ICLR 2025.

selected publications & preprints

(* for equal contribution, see full list here)
  1. dlm-scaling.png
    Revisiting the 16x Pretraining Compute Gap between Diffusion and Autoregressive Language Models
    Guanghan Wang, Bing He, Zhan Shi, Benoit Dumoulin, Hanqing Lu*, and Volodymyr Kuleshov*
    arXiv, 2026
    arXiv Blog Code
  2. d2.png
    d2: Improving Reasoning in Diffusion Language Models via Trajectory Likelihood Estimation
    Guanghan Wang, Gilad Turok, Yair Schiff, Marianne Arriola, and Volodymyr Kuleshov
    ICML, 2026
    arXiv Blog Code
  3. remdm.gif
    Remasking Discrete Diffusion Models with Inference-Time Scaling
    Guanghan Wang*, Yair Schiff*, Subham Sekhar Sahoo, and Volodymyr Kuleshov
    NeurIPS, 2025
    arXiv Video Blog Code
  4. udlm.gif
    Simple Guidance Mechanisms for Discrete Diffusion Models
    Yair Schiff*, Subham Sekhar Sahoo*, Hao Phung*, Guanghan Wang*, Sam Boshar, Hugo Dalla-torre, Bernardo P Almeida, Alexander Rush, Thomas Pierrot, and Volodymyr Kuleshov
    ICLR, 2025
    arXiv Video Blog Code