Yiyang Steven Cai

CS undergrad @ Fudan University, interested in embodied ai and robot learning

[Yiyang Steven Cai @Steven1521426]

🤖 Can robots imitate intent, not just actions?

Excited to present The Imitator Game at #CoRL2026!

📍 Poster #88 | JW Marriott, Lone Star (3F)
🕔 Nov 10, 5–6:30 PM

Come chat!

🔗 https://t.co/DZ1MKiwr2B

[Xunzhe Zhou (周训哲) @XunzheZhou]

We’re releasing The Imitator Game, a hierarchical benchmark for Human→Robot video imitation.

@GeneralistAI @SkildAI showed that scaling human-video-prompting can enable one-shot robot task learning.

We built the benchmark for the next question: How far can robots go beyond copying human actions — toward imitating human intent?

We build a four-level benchmark: L0 scene-identical execution → L1 spatial adaptation → L2 visual/physical generalization → L3 intent-level functional transfer.

Along with it, we release: 20K+ paired real/sim episodes dataset, 200+ L0-L3 task variants and simulation environments, 9 evaluted models / 15 variants, human evaluation Imitator Arena.

Website: https://t.co/tFSPFLAmaF
Paper: https://t.co/uuga029w69
GitHub: https://t.co/2QeVkEJUiD

[Yiyang Steven Cai @Steven1521426]

@zucan_lv 😙😙

[Yiyang Steven Cai @Steven1521426]

Thanks, Sherry! Really excited to see our work out! 🚀

I believe human experience has much more to offer robots than actions to imitate. The challenge is turning that experience into useful guidance for embodied agents.

It’s been a great journey working with this amazing team!

[Sherry Yang @sherryyangML]

Why should agentic robots rediscover what humans already know?

Excited to share our new work, MoCaP (mobile Code-As-Policy), which turns passive human videos of everyday life into a head start for agentic robots: where things are and how to approach them.

Human experience → less robot trial and error.

[Tatsuya Kamijo @tatsukamijo]

Coding agents are emerging as general robot policies, but they are still inefficient (e.g., requiring long real-world exploration), especially in mobile manipulation.

Humans become efficient by watching others.
Can coding agents learn from passive human videos, just like we do?

Introducing MoCaP: Mobile Code-as-Policies for Zero-Shot Mobile Manipulation from Passive Human Videos

↓🧵