Hyunchul Kim

[Hyunchul Kim @kimhynchl]

For more information, check our demo paper or contact us! pdf : https://t.co/VdHyQugb9o

*Additionally, special thanks to Jihoon Shin and Jiho Kang, who helped us capture the motion sets!

[Hyunchul Kim @kimhynchl]

We presented 'Situately', an embodied conversational agent with LLM-powered nonverbal behavior and object interaction, as a research demo at IEEE VR 2026 (Daegu, South Korea)! Thanks to all the participants who actively engaged and enjoyed conversations with Dr. Virtual!

[Hyunchul Kim @kimhynchl]

In this project, Kwang Bin Lee implemented object interaction, Jungsoo Lee designed the environment and scenario, and Sihyun Jeong fine-tuned the language model to align nonverbal behaviors with human judgments, under the supervision of Prof. Sung-Hee Lee and Prof. Jeongmi Lee.

[Hyunchul Kim @kimhynchl]

Attendees noted low latency and high responsiveness, but also reported limitations in turn-taking and motion quality — all valuable feedback. We totally agree, and we are working toward more fluent social interaction with the agent.

[Hyunchul Kim @kimhynchl]

In this work, we leveraged LLM for nonverbal behavior and action planning. To support this, we built a framework consisting of verbal conversation, conversation memory, nonverbal behavior control, and object interaction control — all concurrently coordinated.

[Hyunchul Kim @kimhynchl]

If you are more interested in this work, check out the presentation schedule!

📄 Paper: https://t.co/sYGqWtvETN
🗓 Presentation: April 30 (Wed), 2:10–2:22 PM
📍 Location: G403

[Hyunchul Kim @kimhynchl]

Thrilled to present our paper,
"Quantifying Social Connection with Verbal and Non-Verbal Behaviors in VR Conversations" at CHI 2025!

[Hyunchul Kim @kimhynchl]

We explored how turn-taking, gaze, and nodding behaviors reflect perceived social connection during dyadic VR conversations. Our findings provide a foundation for quantitatively understanding social interactions beyond subjective impressions.

[Fiona Ryan @fionakryan]

Introducing Gaze-LLE, a new model for gaze target estimation built on top of a frozen visual foundation model!

Gaze-LLE achieves SOTA results on multiple benchmarks while learning minimal parameters, and shows strong generalization

paper: https://t.co/Is2NgrrurO

[James Merrill @toThePixel]

You should know about Poisson Disk Sampling. Once you know what it is, you'll spot it in some of the best generative artist's work.

Read more ↩️

[Larry Engineer @larry0x]

Genius 🧠

[microscopic images. @microscopicture]

single neuron trying to connect with the network

[Golan Levin @golan]

I'm just blown away by the amazing performance, features, and easy integration of handsfree.js by @Metamoar (built on TensorFlow & MediaPipe)—with multiple face, hand, & body tracking at 20-60fps in @p5xjs. Here's my example project for tomorrow's class: https://t.co/1EGlyoT6k9

[Keijiro Takahashi @_kzr]

Testing the AR video recorder/player with an outdoor environment. https://t.co/AzV6Y0ifgG

[Byungjoo Lee @yonseiesports]

Yay! A paper got accepted to ACM MM'21 (with Injung Lee and Hyunchul Kim). The title is "Automated Playtesting with a Cognitive Model of Sensorimotor Coordination".

[Hyunchul Kim @kimhynchl]

https://t.co/Fr6HJe1fWt

[zach lieberman @zachlieberman]

Shape grammars / Janos maroscheck

[algoritmic @algoritmic]

Dune Field Simulator https://t.co/fsEl2rT7aq

[Journey @atmostbeautiful]

새 유전자는 어떻게 생길까. 그전까지는 기존 유전자가 복제 과정에서 오류 일으켜 뒤섞이거나 파손되며 생긴다고 봤다. 사실은 새로운 유전자 생성도 허다하다. 환경 적응 위해 유전체 중 쓸모 없었던 부분에서 새로 만들어내기도 한다. 유기체의 진화는 생각 이상으로 창의적https://t.co/7zGTFsaSsJ

[hardmaru @hardmaru]

Privacy and security guide for macOS

This guide looks quite comprehensive. Lots of useful information here.

https://t.co/ijZc9Y2oBP