Hao Liu

I am a research scientist at Google DeepMind. I am also an Assistant Professor of Machine Learning at Carnegie Mellon.

Previously, I completed my Ph.D. in Computer Science at Berkeley, working with Pieter Abbeel. I also spent two years part-time at Google as part of the Brain team.

My research interests focus on foundational problems in deep learning, neural networks, learning objectives, and solving superintelligence.

Native multimodal world model on millions tokens length (paper'24,code).

Ring Attention, Blockwise Transformer (paper'24,paper'23,code).

ElasticTok: adapts computation based on available information in images/videos (paper'24,code).

Open language models: Koala, OpenLLaMa (blog'23,models).

Agentic Transformer for learning decision-making from trial-and-error sequences at scale (paper'23).

Reinforcement learning from unsupervised exploration (paper'22,paper'22,paper'21).