Stone Tao

I am a 3rd-year PhD student at UC San Diego advised by professor Hao Su and am also working at NVIDIA research as a research intern. I'm grateful to have my research be funded by the NSF Graduate Research Fellowship.

My current research interests revolve around advancing embodied AI and robot foundation models via compute-scalable synthetic data like simulation or world models, as well as machine learning tools like RL. The phrase "simulation integrated machine learning" captures much of my main research interests. In the context of machine learning tools for robotics, we can modify both the data generator and ML algorithms to address downstream problems from better sim2real RL to scalable simulation evals of real robot policies. If you are interested in working with me / just want to chat about stuff feel free to reach out to my email / twitter!

On the side, I'm also interested in building and running high-quality open-source AI competitions for education and research. I co-founded the Lux AI Challenge and collaborate frequently with Kaggle to build and deliver these AI competitions.

Previously I was an undergraduate at UC San Diego, advised by Hao Su (on robotics and reinforcement learning) and Steven Dow (on AI x HCI). During my undergraduate I interned at QuantCo (machine learning) advised by Ben Thompson and LaunchDarkly (full-stack software engineering).

News

• September, 2025: One paper (improving massively parallel on-policy RL via reset staggering) is accepted to NeurIPS'25, see you in San Diego!
• March, 2025: ManiSkill3 is accepted at RSS'25 and as an oral presentation at the ICLR Robot Learning Workshop, see you in LA (RSS) and Singapore (ICLR)!
• January, 2025: Two papers accepted to ICLR'25.
• December, 2024: Lux AI Season 3 has started! Compete for a $50,000 prize pool on Kaggle.
• May, 2024: ManiSkill3 Beta has been released! State-of-the-art fully open-source robotics simulation with the fastest parallel visual simulation, faster than most other simulators. 1000s of FPS for visual RL on just google colab! Comes with a ton of features for teleoperation, rendering, reinforcement learning and more.
• May, 2023: The NeurIPS edition of Lux AI Challenge Season 2 focusing on large-scale multi-agent RL was accepted to NeurIPS'23 as an Competition!
• March, 2023: Received the NSF Graduate Research Fellowship!
• March, 2023: Incoming PhD at UC San Diego advised by Hao Su!
• February, 2023: Podcast with Kaggle on Lux AI Season 2 and Reinforcement Learning
• January, 2023: Lux AI Season 2 Started: Check it out!

Publications / Preprints

Papers sorted by recency, Representative papers are highlighted.

Multi-Stage Manipulation with Demonstration-Augmented Reward, Policy and World Model Learning

Adrià López Escoriza, Nicklas Hansen, Stone Tao, Tongzhou Mu, Hao Su

ICML 2025

arXiv / project page / code

Do Vision-Language Models Have Internal World Models? Towards an Atomic Evaluation

Qiyue Gao*, Xinyu Pi*, Kevin Liu, Junrong Chen, Ruolan Yang, Xinqi Huang, Xinyu Fang, Lu Sun, Gautham Kishore, Bo Ai, Stone Tao, Mengyang Liu, Jiaxi Yang, Chao-Jung Lai, Chuanyang Jin, Jiannan Xiang, Benhao Huang, Zeming Chen, David Danks, Hao Su, Tianmin Shu, Ziqiao Ma, Lianhui Qin, Zhiting Hu.

ACL 2025

arXiv / project page / code

ManiSkill-HAB: A Benchmark for Low-Level Manipulation in Home Rearrangement Tasks

Arth Shukla, Stone Tao, Hao Su

ICLR 2025

arXiv / project page / code

Policy Decorator: Model-Agnostic Online Refinement for Large Policy Model

Xiu Yuan*, Tongzhou Mu*, Stone Tao, Yunhao Fang, Michael Zhang, Hao Su

ICLR 2025

arXiv / project page / code

ManiSkill3: GPU Parallelized Robotics Simulation and Rendering for Generalizable Embodied AI

Stone Tao, Fanbo Xiang, Arth Shukla, Yuzhe Qin, Xander Hinrichsen, Xiaodi Yuan, Chen Bao, Xinsong Lin, Yulin Liu, Tse-kai Chan, Yuan Gao, Xuanlin Li, Tongzhou Mu, Nan Xiao, Arnav Gurha, Viswesh Nagaswamy Rajesh, Yong Woo Choi, Yen-Ru Chen, Zhiao Huang, Roberto Calandra, Rui Chen, Shan Luo, Hao Su

RSS 2025
ICLR 2025 Robot Learning Workshop, Oral Presentation

arXiv / project page / code

Lux AI Season 3: Lux AI Season 3: Multi-Agent Meta Learning at Scale

Stone Tao, Akarsh Kumar, Bovard Doerschuk-Tiberi, Isabelle Pan, Addison Howard, Hao Su

NeurIPS 2024 (Competitions Track)

/ project page / code

RFCL: Reverse Forward Curriculum Learning for Extreme Sample and Demonstration Efficiency in RL

Stone Tao, Arth Shukla, Tsekai Chen, Hao Su

ICLR 2024

arXiv / project page / code

MetaWriter: Exploring the Potential and Perils of AI Writing Support in Scientific Peer Review

Lu Sun, Stone Tao, Junjie Hu, Steven Dow

CSCW 2024

arXiv

Lux AI Season 2 (NeurIPS Edition)

Stone Tao, Qimai Li, Yuhao Jiang, Jiaxin Chen, Xiaolong Zhu, Bovard Doerschuk-Tiberi, Isabelle Pan, Addison Howard

NeurIPS 2023 (Competitions Track)

/ project page / code

Abstract-to-Executable Trajectory Translation for One-Shot Task Generalization

Stone Tao, Xiaochen Li, Tongzhou Mu, Zhiao Huang, Yuzhe Qin, and Hao Su.

ICML 2023

arXiv / project page / code

ManiSkill2: A Unified Benchmark for Generalizable Manipulation Skills

Jiayuan Gu†, Fanbo Xiang†, Xuanlin Li*, Zhan Ling*, Xiqiang Liu*, Tongzhou Mu*, Yihe Tang*, Stone Tao*, Xinyue Wei*, Yunchao Yao*, Xiaodi Yuan, Pengwei Xie, Zhiao Huang, Rui Chen, Hao Su.

ICLR 2023

arXiv / project page / code

Emergent Collective Intelligence from Massive-Agent Cooperation and Competition

Hanmo Chen*, Stone Tao*, Jiaxin Chen, Weihan Shen, Xihui Li, Sikai Cheng, Xiaolong Zhu, Xiu Li

NeurIPS 2022 Deep RL Workshop

arXiv / code

Maniskill: Generalizable manipulation skill benchmark with large-scale demonstrations

Tongzhou Mu*, Zhan Ling*, Fanbo Xiang*, Derek Cathera Yang*, Xuanlin Li*, Stone Tao, Zhiao Huang, Zhiwei Jia, and Hao Su.

NeurIPS 2021 (Datasets and Benchmark Track)

arXiv / project page / code

Presentations / Talks

Pinned Projects

ManiSkill

2024/05

Python

Robotics

GPU Parallelized Robotics Simulation and Rendering framework. Designed to be flexible, easy to use, and supports sim2real/real2sim and many robot learning baselines.

Lux AI Challenge Season 1

2021/08

Typescript

Python

Season 1 of the Lux AI Challenge, built for novelty, inclusivity, and accessibility, hosted on Kaggle! This competition is step one in our efforts to build the next generation of AI competitions and environments.

Polytomizator

2018/07

p5.js

JS

HTML & CSS

This is a website that auto generates poly art. It makes use of Delaunay triangulation, image filtering, computer vision techniques, and Poisson disk sampling to help auto-generate better low-poly art.

Awards