Kwang-Sung (Kwang) Jun

associate professor at POSTECH. ML theory and recently LLMs! i am a fan of regret minimalism, both in life and research.

[Kwang-Sung (Kwang) Jun @kwangsungjun]

I met my PhD advisor at #icml2026 who now writes ai-free music!

[Juno KIM @junokim_ai]

Our new paper at #ICML2026 studies why simple on-policy preference learning (e.g., online DPO) can outperform their offline counterparts! Joint work with Jihun Yun @jasondeanlee @kwangsungjun Presenting tomorrow (July 8) 5pm, Hall A #4514!

https://t.co/vANMzMrFvV 1/5

[Kwang-Sung (Kwang) Jun @kwangsungjun]

due to traveling constraints, i will present the poster instead of the co-first authors. i hope to see you at the poster!

[Kwang-Sung (Kwang) Jun @kwangsungjun]

#ICML2026 poster presentation today at 2:30pm:
#4512 "Fixed Budget is No Harder Than Fixed Confidence in Best-Arm Identification up to Logarithmic Factors"

Joint work with Houssam's team at Meta and my team from U of Arizona with co-first authors Kapilan and Yinan

[Kwang-Sung (Kwang) Jun @kwangsungjun]

Now, one can see that the existence of our algorithm and guarantee implies that FB is no harder than FC, ignoring logarithmic factors..!

[Kwang-Sung (Kwang) Jun @kwangsungjun]

@avt_im but yeah! that link is a good starting point for me.

[Kwang-Sung (Kwang) Jun @kwangsungjun]

@avt_im I was intrigued by your remark that with your macro your source looks more like a markdown! Just curious, because the readability of the tex source still often bothers me.

[Kwang-Sung (Kwang) Jun @kwangsungjun]

@avt_im any plans to release your tex macro package?

[Kwang-Sung (Kwang) Jun @kwangsungjun]

@DimitrisPapail “theory is just a hint.” I liked what manfred said before. by the way, the standard generalization theory is still worst case base and focused on the large sample regime. i believe we can get closer by instance dependent analysis and studying data poor regime!

[Kwang-Sung (Kwang) Jun @kwangsungjun]

@DimitrisPapail @madiator what is ‘single batch inf calls’?

[Kwang-Sung (Kwang) Jun @kwangsungjun]

@DimitrisPapail this is the core cs skill -- identifying such a nontrivial bug and finding a workaround. use binary search and write test code and use debugging tools. have the scientist mindset to not rule out anything with no strong evidence..! high respect to those who did this.

[Kwang-Sung (Kwang) Jun @kwangsungjun]

🚀Exciting news! Our project "Orchestrating Large Language Models via Contextual Partial Monitoring" has been selected for the NVIDIA Academic Grant Program! Many thanks to to NVIDIA the hardware support and I look forward to making breakthroughs! #NVIDIAGrant @NVIDIAAIDev

[Kwang-Sung (Kwang) Jun @kwangsungjun]

@Kangwook_Lee @UWMadison @UWMadisonECE congrats again!! well deserved! looking forward to your next journey!

[Kwang-Sung (Kwang) Jun @kwangsungjun]

@SharonYixuanLi congratulations!

[Kwang-Sung (Kwang) Jun @kwangsungjun]

@captain8299 congrats!!

[Kwang-Sung (Kwang) Jun @kwangsungjun]

I am heading to JSM2025, Sunday-Tuesday. #JSM2025 I will give a talk at "Advances in Adaptive Experimentation and Anytime-Valid Inference" session on Sunday. Let's meet up if you like to discuss anything!

[lalit @stochasticlalit]

It was amazing to be part of this effort. Huge shout out to the team, and all the incredible pre-training and post-training efforts that ensure Gemini is the leading frontier model!

https://t.co/KUHpY8y1qL