Junhyuk So

Ph.D candidate @ POSTECH
Area : Efficient ML, Generative Model

[Junhyuk So @goqh_so]

@_TobiasLee Honestly i think this is very good materials for early year students :)

[Junhyuk So @goqh_so]

@HuggingPapers GEPA?

[Junhyuk So @goqh_so]

@LayneHuangLH its very cool work!!

[Junhyuk So @goqh_so]

@LumaLabsAI Is there any papers on architecture details?

[adn 👣 @adn_twitts]

@iclr_conf where are you???

[Junhyuk So @goqh_so]

@mblondel_ml This gives some explanations to why logit(feature) based Speculative decoding works well, like EAGLE.

[Ziwei Liu @liuziwei7]

🗺️Towards A Roadmap of Video World Model🗺️

We conceptualize the progression of video generation through **four generations**, in which the core capabilities ultimately culminating in a world model.

🏡 https://t.co/Cr9T4Uq7QP
📜 https://t.co/NvXmw6JbbG
🧑‍💻 https://t.co/lv4QuAW08m

[Yutong (Kelly) He @electronickale]

Diffusion/Flow-based models can sample in 1-2 steps now 👍 But likelihood? Still requires 100-1000 NFEs (even for these fast models) 😭

We fix this! Introducing F2D2: simultaneous fast sampling AND fast likelihood via joint flow map distillation.

https://t.co/FFfqWnLIwu
1/🧵

[Junhyuk So @goqh_so]

@elonmusk @StefanoErmon @_inception_ai wow elon

[Junhyuk So @goqh_so]

@rosinality Oh i understand the point. I think it's promising because modeling difficulty of visual token space is not that hard

[Oscar Davis @osclsd]

Introducing Generalised Flow Maps 🎉

A stable, few-step generative model on Riemannian manifolds 🪩

📚 Read it at: https://t.co/iCTHedwCxf

💾 Code: https://t.co/MeukcthFN2

@msalbergo @nmboffi @mmbronstein @bose_joey

[Junhyuk So @goqh_so]

@scott_linderman @NeurIPSConf This is really cool! Im also very interested in accelerating AR sampling or Discrete diffusion sampling with newton

[Junhyuk So @goqh_so]

@giffmana There is no free lunch for sure.. But actually we can reduce time complexity with cost of space complexity(or more parallel computation), without sacrificing accuracy..

[Chieh-Hsin (Jesse) Lai ✈️ ICML @JCJesseLai]

Tired to go back to the original papers again and again? Our monograph: a systematic and fundamental recipe you can rely on!

📘 We’re excited to release 《The Principles of Diffusion Models》— with @DrYangSong, @gimdong58085414, @mittu1204, and @StefanoErmon.

It traces the core ideas that shaped diffusion modeling and explains how today’s models work, why they work, and where they’re heading.

🧵You’ll find the link and a few highlights in the thread.
We’d love to hear your thoughts and join some discussions!

⚡ Stay tuned for our markdown version, where you can drop your comments!

[Junhyuk So @goqh_so]

@mpark1999 Cool! Check our Self-Guidance (https://t.co/bR3syunwri) also

[Junhyuk So @goqh_so]

@UnderGroundJeg Very cool work!! Looking forward to the code

[Chenhao Li @breadli428]

People: robots are replacing humanity.

Real: robot dance show at a leading robotic conference without moving a toe.

[Junhyuk So @goqh_so]

@Saba_A96 Very cool work!!

[Junhyuk So @goqh_so]

@mahmoudnafifi @ICCVConference Did you solve the problem? I have same issue and they send me email that my submission is not received to cps system..