Twitxy Saved

Zechen Bai

Verified account This account is verified. Learn more

@ZechenBai

CS PhD @NUSingapore ; Previous Meta MSL, AWS, ByteDance; I build multimodal foundation models and agents. I care about data, eval, post-train, and harness.
Singapore baizechen.site Joined July 2020
492 Following 429 Followers 141 Posts

Latest posts 5

Pinned post

Zechen Bai Verified account @ZechenBai
[1/8] 🦾 Could a strong robot policy already be hiding inside your VLM, not just GPT Astra, but also Gemini, Claude, even light-weight Qwen and InternVL?
🕹️ We are thrilled to introduce Show-Harness, an Embodied Harness that lets VLMs play real robots—like Claude Code for robotics.
No more calibration. No extra VLA.
1:50
15 replies 48 reposts 306 likes 33K views View on X
Zechen Bai Verified account @ZechenBai
Curious about Agentic Robot Control? 🤖

How is it related to multimodal agents?
What really separates Robot Use from Computer Use?
And where do VLMs, VLAs, planning, tools, and embodied interaction fit together?
Check out our latest survey — a map of this fast-moving space 👇
Anno Yanzhe Chen Verified account @Anno_YanzheChen Sep 11
[1/5] Yesterday, we released Show-Harness — and saw frontier VLMs like GPT-6 Astra directly control real robots with striking zero-shot performance. 🦾

Today, we zoom out:
What really changes when multimodal agents move from computer-use to robot-use?

Our new survey on Multimodal Embodied Agents 🧵👇
View quoted post on X
Zechen Bai Verified account @ZechenBai
Harness matters!
Simon Smith Verified account @_simonsmith Sep 10
Very cool, and reminds me of what we saw with ARC-AGI 3. Harness matters. We saw this for coding. We saw this for ARC. Now we're seeing it for robots. GPT-6 Astra can control robots out of the box, but it's a big model, so latency is an issue. But now with a harness like this smaller, faster models can also use robots.
View quoted post on X
Zechen Bai Verified account @ZechenBai
Check out the impressive multimodal agentic capability of the latest Muse Spark model!!
Jiahui Yu Verified account @jiahuiyu Jul 9
Muse Spark 1.1 is here! Available today via API and on meta.ai.

Of all the agentic use cases, I’m most excited about this one: going from a single video shot straight to task completion. 🚀

We’ve also polished and launched Computer Use and Video-Audio Perception for Muse Spark 1.1. Read more: ai.meta.com/blog/introduci…
View quoted post on X