Chuang Gan (@gan_chuang)

Pinned

C

Chuang Gan@gan_chuang · Jun 20

World Simulator, reimagined — now alive with humans, robots, and their vibrant society unfolding in 3D real-world geospatial scenes across the globe! 🚀 One day soon, humans and robots will co-exist in the same world. To prepare, we must address: 1️⃣ How can robots cooperate or…

10

58

243

98

63.0K

C

Chuang Gan@gan_chuang · Jul 22

Spatial reasoning from a single image is inherently difficult, but it becomes significantly easier when leveraging a controlled world model, analogous to the mental models used by humans! Code: github.com/UMass-Embodied…

YYuncong Yang@YuncongYY · Jul 21

Test-time scaling nailed code & math—next stop: the real 3D world. 🌍 MindJourney pairs any VLM with a video-diffusion World Model, letting it explore an imagined scene before answering. One frame becomes a tour—and the tour leads to new SOTA in spatial reasoning. 🚀 🧵1/

1

11

92

37

14.0K

C

Chuang Gan@gan_chuang · Jul 18

Professor Zhao 👍👍👍

WWenting Zhao@wzhao_nlp · Jul 18

I'll be around the ICML venue this afternoon. Message me if you want to meet! These days, I think about reasoning and RL. Also happy to talk about academia vs. industry (I think the lack of compute in academia is a feature not a bug), faculty and PhD student recruiting at UMass.

1

0

13

0

3.0K

C

Chuang Gan@gan_chuang · Jul 15

Excited to be at ICML to present four papers and recruit new faculty for UMass Amherst! We're hiring in generative AI, NLP, and 3D vision—please feel free to reach out if you're interested!

0

2

43

1

4.0K

C

Chuang Gan@gan_chuang · Jul 13

Thank you to AK for introducing our new work on Fast 3D Language Gaussian Splatting! Please try our code: github.com/ZhaoYujie2002/…

AAK@_akhaliq · Jul 11

LangSplatV2 High-dimensional 3D Language Gaussian Splatting with 450+ FPS

2

9

53

19

21.0K

C

Chuang Gan@gan_chuang · Jul 2

Building a World Simulator is my best bet for achieving embodied AGI! I'm truly inspired and grateful to see the next generation of robotics leaders — @zhou_xian_, @theo_gervet, @Zhenjia_Xu, @johnsonwang0810, @yilingq97, and many others — boldly carrying this vision forward and…

GGenesis AI@gs_ai_ · Jul 1

Today, We’re launching Genesis AI — a global physical AI lab and full-stack robotics company — to build generalist robots and unlock unlimited physical labor. We’re backed by $105M in seed funding from @EclipseVentures, @khoslaventures, @Bpifrance, HSG, and visionaries…

0

6

69

8

8.0K

C

Chuang Gan@gan_chuang · Jul 1

🧠 LLMs think too much—and waste tokens! Can we precisely control how long they reason? Introducing Budget Guidance — a thinking-budget-conditioned generation method that controls how long an LLM thinks! We use a lightweight predictor to estimate the remaining reasoning…

gan_chuang's tweet image. 🧠 LLMs think too much—and waste tokens! Can we precisely control how long they reason?

Introducing Budget Guidance — a thinking-budget-conditioned generation method that controls how long an LLM thinks!

We use a lightweight predictor to estimate the remaining reasoning…

10

19

118

76

9.0K

C

Chuang Gan@gan_chuang · Jun 27

VLM can think visually without generating pixels! VLM can think visually without generating pixels! VLM can think visually without generating pixels! 📢 We introduce Machine Mental Imagery (Mirage): a new framework that enables VLM to imagine using latent visual…

gan_chuang's tweet image. VLM can think visually without generating pixels!
VLM can think visually without generating pixels!
VLM can think visually without generating pixels!

📢 We introduce Machine Mental Imagery (Mirage): a new framework that enables VLM to imagine using latent visual…

8

132

743

523

77.0K

C

Chuang Gan@gan_chuang · Jun 21

Attending RSS for the first time and giving a talk tomorrow at the Learning Structured World Models for Robotic Manipulation workshop! At midnight, I made a last-minute crazy decision to change my talk content to Virtual Community — to honor the incredible hard work of my…

CChuang Gan@gan_chuang · Jun 20

World Simulator, reimagined — now alive with humans, robots, and their vibrant society unfolding in 3D real-world geospatial scenes across the globe! 🚀 One day soon, humans and robots will co-exist in the same world. To prepare, we must address: 1️⃣ How can robots cooperate or…

2

9

63

9

6.0K

C

Chuang Gan@gan_chuang · Jun 20

Digital twin of (the future of) our physical world?

CChuang Gan@gan_chuang · Jun 20

World Simulator, reimagined — now alive with humans, robots, and their vibrant society unfolding in 3D real-world geospatial scenes across the globe! 🚀 One day soon, humans and robots will co-exist in the same world. To prepare, we must address: 1️⃣ How can robots cooperate or…

0

3

11

1

3.0K

C

Chuang Gan@gan_chuang · Jun 20

Wow, this is so cool! Have been dreaming of building agents that can interact with humans via language communications, and the world via physical interaction (locomotion, manipulation, etc). Definitely a great step-stone and playground!

CChuang Gan@gan_chuang · Jun 20

World Simulator, reimagined — now alive with humans, robots, and their vibrant society unfolding in 3D real-world geospatial scenes across the globe! 🚀 One day soon, humans and robots will co-exist in the same world. To prepare, we must address: 1️⃣ How can robots cooperate or…

0

3

12

3

3.0K

C

Chuang Gan@gan_chuang · Jun 20

guys, real geospatial data is a total goldmine for digital agents. step away from the web browser and get real. (we explored a bit in virl-platform.github.io, but building a simulation-ready pipeline like this could take things way further)

CChuang Gan@gan_chuang · Jun 20

Virtual Community provides an online pipeline that automatically generates 3D scenes from real geospatial data, performing comprehensive cleaning and enhancement of both geometry and texture — including mesh simplification, texture refinement, object placement, and automatic…

4

19

104

41

20.0K

C

Chuang Gan@gan_chuang · Jun 17

🤖Can world models quickly adapt to new environments with just a few interactions? Introducing AdaWorld 🌍 — a new approach to learning world models conditioned on continuous latent actions extracted from videos via self-supervision! It enables rapid adaptation, efficient…

5

37

159

116

25.0K