Feng Chen

Ph.D. student. at the University of Hong Kong, advised by prof. Yi Ma
Graduated from IIIS Tsinghua University 2024 (Yao Class)

[Feng Chen @Winniechen02]

🤖Can robots play Texas Hold’em? 🤔

We introduce DexHoldem, a real-world benchmark for dexterous manipulation and embodied agents.

After fine-tuning leading VLAs and pairing them with frontier agents, we found:

Robots can move the chips — sometimes.
Agents can read the table state — sometimes.

Unfortunately, poker requires both to happen correctly in the same hand.

So no, robots are not ready to take your money at the casino — yet.

More below 👇

[Ilya Sutskever @ilyasut]

if you value intelligence above all other human qualities, you’re gonna have a bad time

[Feng Chen @Winniechen02]

@simon_zhai @Polymarket Then how many cups you want to drink?

[Feng Chen @Winniechen02]

If what the reviewer, area chair, meta review and program chair do is just feeding all (or even all) information into AI, what is the point of my submitting the manuscript to the meeting?
Then I congratulate neurips on becoming the first thing in the world to achieve RSI!

#Neurips2026 #RSI

[Feng Chen @Winniechen02]

@docmilanfar Not that useful

[Feng Chen @Winniechen02]

Let’s put them side by side. 👀
Right: A month ago, GPT-5.6 Sol barely managed to complete the two-disk task—even with a whole toolbox of helpers.
Left: GPT-6 made it halfway through the five-disk task, using only rendered camera images and robot joint angles as observations.
Why only halfway? You’ll have to ask Astra why it’s so hungry for tokens. 😂
We’re releasing a repo soon with our latest experiments on LLMs + harnesses directly controlling robot arms and dexterous hand across several simulation environments.
Once it’s public, bring your own LLM and harness and give it a spin! We’d love to see what you build—and how far you get before the tokens run out. 🤖
Stay tuned, and happy experimenting!
If Tibo could help me get some reset, the release would be even faster! @thsottiaux

[Yann LeCun @ylecun]

Looks like we have over 500 submissions to ICLR this year.

https://t.co/jiawnKtnIF

[Feng Chen @Winniechen02]

@TianzheC Why Chen Xu don’t want to study under your supervision, you need to be a better supervisor!🤣🤣🤣

[Feng Chen @Winniechen02]

For me, learning how to do something is more important than make something work. Academia is the place for students to learn, not the place to work for supervisors’s ridiculous requirements.

[Toru @ToruO_O]

Inspired by recent thoughts from senior researchers whom I deeply respect (e.g. @JitendraMalikCV @Michael_J_Black @Ken_Goldberg @phillip_isola), I wrote down some thoughts as a junior researcher too.

The recent Astra demos in dexterous manipulation are truly impressive. But they also raise two questions that I don’t think academia has good answers to yet:
- How should credit be assigned when an agent synthesizes many prior works into a new research result?
- How can academia attract and retain talents when so much of its incentive structure relies on credit?

I reflect on these questions in more detail in a blog post. Curious to hear what others think, and debates are very welcome!

https://t.co/52qCV6CPMY

[Feng Chen @Winniechen02]

@simon_zhai @ToruO_O @JitendraMalikCV @Michael_J_Black @Ken_Goldberg @phillip_isola Why simon doesn’t write one? I want to see your blog😋

[Feng Chen @Winniechen02]

I could estimate GPT will do all these skills but I don’t know it could be so fast!

[Jianglong Ye @jianglong_ye]

Evolving dexterity with GPT-6 Astra 🖐️

Been trying Astra recently. Its zero-shot dexterous manipulation is already quite surprising. More interesting is seeing it learn and improve through simulation training, from pen spinning and Rubik's Cube to hammer use.

The real goal would be to evolve this dexterity in the real world.

https://t.co/GZjQTsklYc

[Feng Chen @Winniechen02]

@ToruO_O Then maybe it would be a good idea to have a try! Maybe it could be a new project or direction for frontier models 🤣

[Feng Chen @Winniechen02]

@ZeYanjie ze哥花了多少token多久跑出来的

[Feng Chen @Winniechen02]

29hours, GPT6, only vision input and perception state, 38 times trying, rotating the pocket cube one time.

Something interesting is GPT6 tried to access more information in simulator but detected by itself and killed.

[Feng Chen @Winniechen02]

@Quaocchoo Yeap, but it would be solved faster than you guess.

[Feng Chen @Winniechen02]

The recent attempt in agentic robotics has inspired me.

1. The grasping task usually has a good effect, because the grasping itself is closer to the processing of static scenes (including dexterous hand grasping!)
2. Collisions and avoiding collisions are difficult for LLM, and it is difficult to solve by yourself without human help (humans have some mobile experience to prevent collisions)
3. At present, dexterity operation is still the territory of imitation learning and reinforcement learning, but who knows how long it will last?

Final question, in addition to collecting data, what skills do you currently master within five years cannot be replaced by LLM?

Thanks Pei Zhou for the interesting demo. Thanks @alex_kai2020 @Lixuan_thu for communicating and helping.

[Feng Chen @Winniechen02]

@RamnathAbhishek Yes, they’re tools but hard to say they’re good tools🤣
But I believe it would be better as a flying wheel.

[Feng Chen @Winniechen02]

Tips: this demo is not from GPT6, it just from 5.6sol! So don’t be too surprised about models capabilities.

[Feng Chen @Winniechen02]

@TianzheC Learning from true master on LLM🤣