Junwon Seo

PhD Student @CMU_Robotics

[Junwon Seo @Junwon__Seo]

Video world model imaginations🌎💭can miss critical but plausible outcomes of robot actions.

Introducing 𝙎𝙩𝙧𝙚𝙨𝙨𝘿𝙧𝙚𝙖𝙢: inference-time steering for video WMs, imagining plausible✅, high-impact⚠️ futures for 𝙧𝙤𝙗𝙪𝙨𝙩🛡️ policy evaluation and improvement.

(1/15)

[Andrea Bajcsy @andrea_bajcsy]

*Very* impressive work from Junwon--he is able to generate synthetic but plausible rare events from a pre-trained video world model (WM): spilling beans during pouring, near-misses during driving, etc. One step closer to doing robust policy evaluation and improvement in WMs!

[Junwon Seo @Junwon__Seo]

@vai_viswanathan Thanks for the wonderful summary!

[Junwon Seo @Junwon__Seo]

This work was 𝒉𝒂𝒑𝒑𝒊𝒍𝒚 done in collaboration with the Autonomous Vehicle Research Group at NVIDIA, with @Veer_Sushant, @tianran_, @wenhaoding95, @apoorva__sharma, Karen Leung, Edward Schmerling, and @drmapavone, under the guidance of @andrea_bajcsy.

(15/15)

[Junwon Seo @Junwon__Seo]

Video world model imaginations🌎💭can miss critical but plausible outcomes of robot actions.

Introducing 𝙎𝙩𝙧𝙚𝙨𝙨𝘿𝙧𝙚𝙖𝙢: inference-time steering for video WMs, imagining plausible✅, high-impact⚠️ futures for 𝙧𝙤𝙗𝙪𝙨𝙩🛡️ policy evaluation and improvement.

(1/15)

[Junwon Seo @Junwon__Seo]

There are many more video results on our interactive website and in the paper. You can also try StressDream with the code:

🌐Project Page: https://t.co/gh8d5L7WiH
📄Paper: https://t.co/ZrPvgvv2Ip
💻Code: https://t.co/9GKo49rRf2

(14/15)

[Hyun Joe @hyunjoej]

Excited to share the first paper of my PhD!

If you’ve ever tried to control a VLA via natural language, you know it rarely does what it is told. 🗣️ We introduce a multi-stage pipeline for training a Language Feedback Policy (LFP) to steer a VLA in-the-loop.

[Yilin Wu @YilinWu11]

So excited to share 𝚆𝙴𝙰𝚅𝙴𝚁 🌎, co-led with @arnavkj95 !

World models are becoming a powerful tool for robot learning — but for real robots, they need to be more than visually realistic. They also need to be consistent, efficient, and useful for decision-making. 🤖✨

[Junwon Seo @Junwon__Seo]

I’ll be at ICRA in Vienna next week to present AnySafe. Feel free to reach out if you’re interested in our work, world models, robot safety, uncertainty quantification, or just chatting about managing life and research!

[Junwon Seo @Junwon__Seo]

[Accepted to ICRA'26]
Is “failure” for robots fixed? Most safe control methods assume safety constraints are known a priori and remain fixed from training to deployment. But what if they change at deployment? AnySafe online adapts a latent-space safe controller at runtime! [1/9]