PhD Student at TU Berlin
Joined February 2026
- Tweets14
- Following67
- Followers61
- Likes59
Gregor Kornhardt retweeted
Join our reading group next Monday!
In DLM research, evaluating sample quality is still an open question for unconditional generation. Happy to see more research on this!
Topic: Measuring Gaps between Text Distributions by Double Slided Wasserstein Metrics
Presenter: Gregor Kornhardt (@gregorkornhardt)
This Monday, Oct 26, Gregor Kornhardt (@gregorkornhardt) will present his paper:
SOL: Measuring Gaps between Text Distributions by Double Sliced Wasserstein Metrics
Recently there has been a lot of discussion about the difficulties of evaluating diffusion language models.
We instead introduce: SOL, a sample based distance between generated and real text distributions. 🧵
Check out the paper!
Website: gregorkornhardt.github.io/SO…
arXiv: arxiv.org/pdf/2610.06513
We also have a package ready, to compare your models.
Code: github.com/GregorKornhardt/S…
This was a super fun collaboration with @JChemseddine, @reayhaatye and Gabriele Steidl.
Some of the discussion that motivated this work: @antoniofrancaib & @AlexanderTong7 in arxiv.org/pdf/2606.08417 and also xitter.cf/Sam_Acqua/status/20545…
Gregor Kornhardt retweeted
Happy to share that our paper was accepted to #NeurIPS2026 as a spotlight! 🎉
A while ago we presented an early version of our work on spherical flows for categorical data at @diffusion_llms. The paper has grown a lot since, so it felt like time for a thread...
Gregor Kornhardt retweeted
A while ago we presented an early version of our work on spherical flows for categorical data at @diffusion_llms. The paper has grown a lot since, so it felt like time for a thread...