Liang Pan (@liangpan

Liang Pan

391 posts

Liang Pan

@liangpan_t

Researching Physical AGI at Light Origins (lightorigins.com/en), PhDing @HKUniveristy

Joined May 2019

Pinned
Liang Pan
@liangpan_t
Mar 29, 2025
This year, an important mission is to let simulated humanoids solve long-term manipulation tasks in challenging dynamic, cluttered environments. This is a simple demonstration achieved by TokenHSI (CVPR 2025). The humanoid here is controlled by a single, unified model.
00:00
150K
Liang Pan
@liangpan_t
Feb 27, 2024
Optimal Reciprocal Collision Avoidance (ORCA) is really amazing! Here is a demo for using it to control 64 simulated humanoid robots.
00:00
9.7K
Liang Pan
@liangpan_t
Apr 3, 2025
Just released long-horizon task completion demo and a pre-trained model. We train the model with a fixed scene init configuration. So changing the init config will probably make the long-term task fail. But never mind, we can fix it by applying randomness during training.
00:00
00:00
00:00
00:17
Liang Pan
@liangpan_t
Mar 29, 2025
This year, an important mission is to let simulated humanoids solve long-term manipulation tasks in challenging dynamic, cluttered environments. This is a simple demonstration achieved by TokenHSI (CVPR 2025). The humanoid here is controlled by a single, unified model.
7.9K
Liang Pan
@liangpan_t
Apr 2, 2025
Thank you for the sharing, AK! Talk is cheap, show me the code. We've released the foundational skill learning part of TokenHSI. Full code, data, and checkpoints are available now at : github.com/liangpan99/Tok… Still working on release more. Stay tuned! #CVPR2025
AK
@_akhaliq
Apr 1, 2025
TokenHSI is out on Hugging Face Unified Synthesis of Physical Human-Scene Interactions through Task Tokenization
00:00
GitHub - liangpan99/TokenHSI: [CVPR 2025 Oral] TokenHSI: Unified Synthesis of Physical Human-Scene...
From github.com
17K
Liang Pan
@liangpan_t
Feb 6, 2024
Finally got some free time to do this. Code for my #3dv2024 physics-based human-scene interaction project is ready. It will be released in next few days!🤩
00:00
6.4K
Liang Pan
@liangpan_t
Aug 21, 2023
Excited to share our new work "InterScene" (arxiv.org/abs/2308.09036). We propose a physics-based framework that enables physically simulated characters to perform long-term interaction tasks in diverse, cluttered, and unseen 3D scenes. Project page: liangpan99.github.io/InterScene/ 1/5
00:00
12K
Liang Pan
@liangpan_t
Jan 24, 2025
A brand-new framework for learning physical character-scene interactions is coming! Some generated motions:
00:00
7.9K
Liang Pan
@liangpan_t
Jan 12, 2024
I would like to say that @xbpeng4's AMP is the most important work for me. Any interesting idea can be quickly validated using AMP with IsaacGym. Here are some demos I made recently: (1) a policy that learns various walking styles (the idea comes from cASE @frankzydou).
00:00
5.4K
Liang Pan
@liangpan_t
Apr 29, 2025
Let me roughly guess the training pipeline. 1. Collect expert demos. 2. Train the base policy using BC. 3. Train the reward model that predicts the task progress. 4. Deploy the policy in real world to collect new data. 5. Label the data with the RM. 6. Offline RL? Any ideas?
Jason Ma
@JasonMa2020
Apr 29, 2025
Introducing Dynamism v1 (DYNA-1) by @DynaRobotics – the first robot foundation model built for round-the-clock, high-throughput dexterous autonomy. Here is a time-lapse video of our model autonomously folding 850+ napkins in a span of 24 hours with • 99.4% success rate — zero
00:00
5K
Liang Pan
@liangpan_t
Jun 13, 2025
🤗Today I'll be presenting our @CVPR oral paper 𝑻𝒐𝒌𝒆𝒏𝑯𝑺𝑰. 𝐎𝐫𝐚𝐥 𝐒𝐞𝐬𝐬𝐢𝐨𝐧 𝟐𝐁: 𝐇𝐮𝐦𝐚𝐧 𝐌𝐨𝐭𝐢𝐨𝐧 📍ExHall A2 🕝Fri 14:15 - 14:30 𝐏𝐨𝐬𝐭𝐞𝐫 𝐒𝐞𝐬𝐬𝐢𝐨𝐧 𝟐 📍ExHall D #159 🕝Fri 16:00 - 18:00 Please check out this thread for a quick preview! 1/8
3K
Liang Pan
@liangpan_t
Mar 29, 2025
Replying to @liangpan_t
We just uploaded all demo videos in the homepage!!🤗 Project page: liangpan99.github.io/TokenHSI/
00:00
3K
Liang Pan
@liangpan_t
Mar 29, 2025
Replying to @liangpan_t
Project page: liangpan99.github.io/TokenHSI/ I'm still working on building it.
4K
Liang Pan
@liangpan_t
Oct 22, 2023
Happy to announce our paper InterScene is accepted to #3DV2024! Check it out in this thread.
Liang Pan
@liangpan_t
Aug 21, 2023
Excited to share our new work "InterScene" (arxiv.org/abs/2308.09036). We propose a physics-based framework that enables physically simulated characters to perform long-term interaction tasks in diverse, cluttered, and unseen 3D scenes. Project page: liangpan99.github.io/InterScene/ 1/5
00:00
4.1K
Liang Pan
@liangpan_t
Mar 5, 2025
When I heard the algorithm that created me (RL) won the Turing Award...
00:00
2.1K