Sohyeon Kim @ ICML

visiting @Stanford
phd-ing @SeoulNatlUni
reasoning with non-parametric memory

[Sohyeon Kim @ ICML @ohmyksh]

If you’re coming to #ICML2026 🇰🇷, this Seoul Guide might be All You Need!

I’ve lived in Seoul for 20+ years and curated this all-in-one guide: Top 5 must-visits, 100+ recs with Google Maps, practical local tips, and more.

Check it out: https://t.co/AdN1sHMg1I

DMs are open and happy to give more details!

[Delta Institute @ ICML @DeltaInstitutes]

🚨 Headed to ICML 2026?

We made a conference-wide Slack with channels for people to find afterparties, meet other attendees, and more!

https://t.co/rCTrHWpnAf

[Sohyeon Kim @ ICML @ohmyksh]

@angli_ai perfect! and I just made a few more improvements :)

[Sohyeon Kim @ ICML @ohmyksh]

If you’re coming to #ICML2026 🇰🇷, this Seoul Guide might be All You Need!

I’ve lived in Seoul for 20+ years and curated this all-in-one guide: Top 5 must-visits, 100+ recs with Google Maps, practical local tips, and more.

Check it out: https://t.co/AdN1sHMg1I

DMs are open and happy to give more details!

[Sohyeon Kim @ ICML @ohmyksh]

@clemissima that’s so lovely to hear :)

[Sohyeon Kim @ ICML @ohmyksh]

@raila_san Thanks for the invite! I’ll have to see if I can make it

[Sohyeon Kim @ ICML @ohmyksh]

Hope there’ll be lots of serendipitous encounters!

maybe we should start an ICML megathread where we share what we’re working on, then grab some good food from my list lol

[Sohyeon Kim @ ICML @ohmyksh]

@shafayat_sheikh thanks! honestly same here, that’s what makes living in Seoul so exciting!

[Sohyeon Kim @ ICML @ohmyksh]

@GuptaManupriya I actually use Naver Maps most of the time. tbh I made this Google Maps version bc I figured it’d be much more familiar for international visitors and easier to access English reviews :)

I’m planning to add a Naver Maps version soon!

[Sohyeon Kim @ ICML @ohmyksh]

@hagsaeng_bag ty! I’m actually sending it to my friends too lol

[Sohyeon Kim @ ICML @ohmyksh]

@soldni happy to hear that!

[Sohyeon Kim @ ICML @ohmyksh]

@beirmug Hope it comes in handy next time you’re in Seoul :)

[Sohyeon Kim @ ICML @ohmyksh]

@shafayat_sheikh Ty! Living in Seoul is pretty awesome

[Sohyeon Kim @ ICML @ohmyksh]

@JangLawrenceK Definitely! just be careful with the spice level at Yupdduk :)

there are a lot more great tteokbokki chains too, I'll add more in an update soon!

[Sohyeon Kim @ ICML @ohmyksh]

@jacobli99 and 10+ desserts too lol

[Yoonho Lee @yoonholeee]

Probably the best Seoul guide you'll find if you're going to ICML. She knows more/better spots than most lifelong Seoul residents

Also, Sohyeon is currently a visiting researcher at Stanford and is closely working with me on retrieval/reasoning; reach out to chat with her!

[Chanwoo Park @chanwoopark20]

Excited to be attending #ICML2026 in Seoul! 🇰🇷

This guide looks very helpful. Even as a Korean, it's hard to know everything in a mega city like Seoul!

I'll be around throughout the conference and would love to connect, especially with researcher in academia and industry working on LLMs and VLAs, particularly science of pretraining and post-training.

Always happy to learn, exchange ideas, and share experiences. I'll also be introducing my ICML papers soon. Stay tuned!

[Yoonho Lee @yoonholeee]

https://t.co/jCgH0doXCQ

[kache @yacineMTB]

you can outsource your thinking
but you cannot outsource your understanding

[Sohyeon Kim @ ICML @ohmyksh]

@aniketthh Cool, I actually read your paper recently! Feels like a nice starting point for this line of work.

[Sohyeon Kim @ ICML @ohmyksh]

Love this idea! Makes me wonder how we retrieve the right parent pair for the next insight from arXiv. It feels like a really challenging IR task, almost like research taste.

[Anikait Singh @Anikait_Singh_]

🚨🚨New Paper: Generative Insight Anticipation from Scientific Literature

Human scientists achieve breakthroughs by "standing on the shoulders of giants," synthesizing profound insights from disparate sources. While LMs show promise in scientific discovery, they currently struggle to reliably generate ideas of true impact, diversity, and feasibility.

Introducing "insight anticipation," a novel task challenging LLMs to generate a downstream paper's core scientific contribution directly from its foundational prior works. We also introduce GIANTS-4B, an RL-trained model that significantly outperforms frontier models at this task! 🧵⬇️