Research

[March 2025]
Scalable Structured Metacognition with Test-Time RL Tree Policies
RL tree search over decoding policies for discrete diffusion models. done as a test-time scaling axis for gemini diffusion. [paper] | [code] † *
[January 2025]
Scaling Inference for Robotics Diffusion Models
variance-aware inference scaling methods for robotics diffusion models. † *
[October 2024]
Trajectory-Search Diffusion Inference Scaling
theoretical framework and proof-of-concept for trajectory-search inference scaling of stochastic diffusion models, expanding on deepmind's work for deterministic models. [paper] | [code] *
[July 2024]
RL-Embedded OS
rl-embedded os. experimented with rust and lower-level systems. [code]
[January 2018]
Computer Vision for Brain Tumor Diagnosis
medical imaging ai using mri images and generative models for dataset balancing. †
† = sole contributor.
* = compute sponsored by senior DeepMind researchers (affiliations under NDA).