University of Cambridge > Talks.cam > Machine Learning Reading Group @ CUED > Diffusion Language Models

  Browse A-Z

Log in

University Account External (via Google)

🔐 Log In

Information on

Subscribing to talksdetails Finding a talkdetails Adding a talkdetails Disseminating talksdetails Help and Documentationdetails

ℹī¸ Information

Diffusion Language Models

Download to your calendar using vCal

If you have a question about this talk, please contact Xianda Sun .

Diffusion Language Models (DLMs) have recently emerged as a promising alternative to the dominant autoregressive paradigm for text generation. Unlike GPT -style models that generate text sequentially, DLMs employ an iterative denoising process that starts from a noisy representation and progressively reconstructs coherent text, enabling parallel token generation and bidirectional context modelling. While early research explored both continuous and discrete diffusion formulations for text, the discrete masked diffusion objectives have recently dominated the landscape, allowing DLMs to effectively scale to larger model sizes and achieve competitive perplexity on standard benchmarks, as evidenced by models such as LLaDA, Mercury, Gemini Diffusion or Seed Diffusion. MDL Ms have attracted a lot of attention for their potential to speed up inference through parallel decoding and improve controllability. In this talk, we will explore the formulation of DLMs, discuss their potential advantages and shortcomings, and dive into the most recent and influential works in this new area.

This talk is part of the Machine Learning Reading Group @ CUED series.

This talk is included in these lists:

Note that ex-directory lists are not shown.

Other lists

Goldsmiths 1, Lecture Theatre, Department of Materials Science & Metallurgy Audit Room, Kings College Cambridge Queens' College, Cambridge

Other talks

Using numerical simulations for solving spectral geometry problems Royal Papworth Hospital and Acute Medicine Seminars in Cancer Fun experiments you can only do with frogs Morning Tea and Coffee Title to be confirmed
 

Š 2006-2025 Talks.cam, University of Cambridge. Contact Us | Help and Documentation | Privacy and Publicity