HOrIc& RO7HrEs#uNl NB% GDjxO2Y0
> PhD at , advised by Caglar Gulcehre
> Now: Student Researcher at Google DeepMind, London
I work on diffusion language models for fast, controllable, and principled alternative to standard LLMs.
Research Focus
- Fast diffusion language models (SDTT, PGM, BlockGen)
- Bridging continuous and discrete diffusion (Duality I, Duality II)
- Improved discrete diffusion samplers (Ψ-Samplers, Loopholing)
Selected Work
Preprint 2026
Language Modeling with Hyperspherical Flows
𝕊-FLM is a hyperspherical flow language model that replaces Gaussian noise on one-hot vectors with rotations on the unit sphere.
ICLR 2026 · oral
Partition Generative Modeling: Masked Modeling Without Masks
Partition Generative Models (PGMs) save compute by never processing [MASK] tokens at inference.
ICLR 2026
The Diffusion Duality, Chapter II: Ψ-Samplers and Efficient Curriculum
Ψ-samplers unlock test-time scaling for uniform discrete diffusion.
Latest News
- Jul. 2026
I started a research internship at Google DeepMind in London!
- Apr. 2026
BlockGen was accepted to the ReALM-GEN workshop at ICLR 2026 and awarded a spotlight talk!
- Jan. 2026
Three papers were accepted to ICLR 2026: Partition Generative Modeling (PGM), Loopholing Discrete Diffusion, and The Diffusion Duality, Chapter II: Ψ-Samplers and Efficient Curriculum. PGM was awarded an oral presentation (top 1.13% of submissions)!
