Richard Gao (@_rwgao) profile picture

@BrownUniversity, model-based reinforcement learning

Providence, RI
Joined July 2023
The real world is full of surprises, like a gust of wind while driving! Intelligent world models must reason about them, and imagine what would have happened under another action... We give JEPA world models this ability with provable guarantees 🧵⬇️ w/ @KerdaVaraku and @ylecun
10
38
3
314
20,595
A huge step towards fast, reproducible and shareable world modeling research! Only store a few states as your dataset, and render high resolution lossless frames on the fly at near zero cost! This is a standalone library, use it with any stack (pytorch/jax/stable-worldmodel/...)
Scared for the future of open source? Black Friday discounts got you disappointed? NO MORE! With stable-envs: - save 99% of disk - rendering speedups of up to 20x faster than baseline - train world models up to 30% faster end-to-end - on 80+ of the most popular envs - focus on training, modeling, and NOT data pipelines! - open source! (see thread)🧵 We'd love to see you use it! If your environment is not supported yet, please open a PR!
1
5
86
3,576
We're releasing an open-source collection of 80 simulation environments, and they're *fast* ⚡ stable-envs train WMs up to 30% faster, and with much better GPU utilization. Check it out! arxiv: arxiv.org/abs/2611.23151
7
24
1
207
8,361
Richard Gao retweeted
Scared for the future of open source? Black Friday discounts got you disappointed? NO MORE! With stable-envs: - save 99% of disk - rendering speedups of up to 20x faster than baseline - train world models up to 30% faster end-to-end - on 80+ of the most popular envs - focus on training, modeling, and NOT data pipelines! - open source! (see thread)🧵 We'd love to see you use it! If your environment is not supported yet, please open a PR!
7
11
2
113
16,851
Richard Gao retweeted
An intelligent machine needs three things: a brain, senses, and hands. We gave AI a brain. We taught it to see. Now we're giving it hands. Models can propose a thousand hypotheses, but breakthroughs come from testing them. Today we're launching @primordiant : general-purpose robot scientists that run experiments at scale. AI mastered bits. Now it's time for ATOMS.
34
52
10
159
46,895
Richard Gao retweeted
Linear Representation or did we just get lucky? New paper: Is a Linear Probe evidence of a Linear Representation? arxiv.org/abs/2610.22655 🧵
Wow, I did not expect that DINOv3's global [CLS] token linearly represents the continuous geometric latents of dSprites (size & X/Y position) 🤯 It only took me 3.5 years to finally run this experiment 😂 I'm looking to do more of this MechInterp work, dissecting foundation models like biological artifacts and building theory. If you want to collaborate (especially students looking for a fun project) reach out! 🔬🤖
6
47
4
506
25,008
The Platonic Representation Hypothesis is now a Theorem ◼️ Train JEPAs separately, on any modalities: they provably learn the same representation up to one rotation. The result? A 109M JEPA-Text model that aligns with vision models better than 72B LLMs. w/ @ylecun 🧵
41
188
30
1,118
86,480
Richard Gao retweeted
Your world model may know less about the world than you think. I've always wondered what world models really learn. Humans don't memorize every trajectory. We learn rules we can reuse. Can world models do the same? We built AnyGoal to find out. hazel-heejeong-nam.github.io…
13
27
4
186
11,141
Introducing H-JEPA: the first end-to-end learned hierarchical world model for long-horizon visual planning! Built atop SIGReg/LeWM, H-JEPA provides a stable training recipe where semantic abstractions naturally emerge across levels... Paper+Code: arxiv.org/abs/2610.06805 🧵👇
19
78
12
590
42,194
Richard Gao retweeted
Blerta and I will be presenting this at COLM on Oct 6! Find the location here: colm.cc/virtual/2026/poster/…
Which state-tracking problems can transformers length generalize on? Transformers' performance on different finite-state tasks varies drastically, even when the tasks are structurally similar. Now, for the first time, we can precisely predict when length generalization occurs.
4
24
842
Richard Gao retweeted
Replying to @typesafeai
@typesafeai turns out Jev can (kind of) play atari too
4
2
4
16
2,903
Richard Gao retweeted
Looks like science is making a comeback! Over 100k impressions on LinkedIn is super encouraging. Anyway, I'm building sth new and ambitious at @ycombinator with my friends @OwenTower1 and @NiuCooper. We are calling it Embodied Scientific Intelligence (ESI) @superradiant_ai Scientific intuition is built through contact with the physical world. We want to help AI bridge the gap between mind and body and focus that intelligence on scientific discovery. It’s a hard problem. But if we crack it … oh man! What an exciting future for humanity!
3
22
61
1,517
Richard Gao retweeted
We’re sharing a solution to the Navier-Stokes Millennium Prize Problem, one of the deepest problems at the frontier of mathematics. The proof was produced by a group of agents, using an OpenAI next-generation model significantly more capable than GPT-6 Astra. The problem concerns whether the description of smooth three-dimensional fluid motion modeled by the Navier-Stokes equations can break down. It has remained unresolved for roughly 90 years.
5,637
19,972
14,813
119,668
75,462,779
Richard Gao retweeted
Which state-tracking problems can transformers length generalize on? Transformers' performance on different finite-state tasks varies drastically, even when the tasks are structurally similar. Now, for the first time, we can precisely predict when length generalization occurs.
7
68
8
474
64,247
Richard Gao retweeted
Officially a Doctor! 🎓🤖 Yesterday I defended my PhD on Scaling Trusted Autonomy: exploring how we can build generally capable robots whose decisions remain verifiable, interpretable, and worthy of our trust. I’ll share more over the coming weeks. For now … I’ll just be watching the sun rise on a slightly more grateful universe. 😄 – at Providence, RI
1:20 AM. Final midnight stretch before my biggest presentation yet. PhD defense tomorrow. Five years of research, distilled into 50 minutes. 🎓🤖
44
91
11
681
73,886
Richard Gao retweeted
1/ I’ve been working on a new open-source project: LeRobot 3D 🤖📷 It’s a 3D-grounded teleoperation stack for the SO101 robot, built to make 3D perception and robot geometry easier to use as part of the LeRobot ecosystem. 🧵👇
2
13
34
7,887
Richard Gao retweeted
Would you like to join the research effort on JEPA and World Models easily? After a full year of hard work, we’re excited to finally release stable-worldmodel: an open-source, scalable platform built to accelerate JEPA & World Model research! 📄: github.com/galilai-group/sta…
42
274
15
1,837
120,618
Richard Gao retweeted
What does JEPA actually learn? We can finally prove it 🌍 So excited to share our theory of identifiable World Models: LeJEPA recovers the latent variables of the world. Plan in the learned World Model as if it were real, same shortest path. 📄: klindtlab.github.io/lejepa-i…
47
256
23
2,139
225,619
Delighted to announce our 3rd world modeling workshop! After NYC and Montreal, we are now headed to Chicago! - August 31st to September 2nd - CfP and details on the website: wm-booth.org - @ylecun and @Diyi_Yang already confirmed, many more to be announced soon!
13
24
3
116
48,558
Richard Gao retweeted
Advanced Machine Intelligence (AMI) is building a new breed of AI systems that understand the world, have persistent memory, can reason and plan, and are controllable and safe. We’ve raised a $1.03B (~€890M) round from global investors who believe in our vision of universally intelligent systems centered on world models. This round is co-led by Cathay Innovation, Greycroft, Hiro Capital, HV Capital, and Bezos Expeditions, along with other investors and angels across the world. We are a growing team of researchers and builders, operating in Paris, New York, Montreal and Singapore from day one. Read more: amilabs.xyz/ AMI - Real world. Real intelligence.
337
852
481
8,396
5,012,883