Dongyeop Kang (DK)

@dongyeopkang

Assistant professor @UMNComputerSci, a member of @MinnesotaNLP group, research on human-centered language technologies, he/him/his

Minneapolis, MN

dykang.github.io

Joined April 2018

213 Following

1,072 Followers

Replies Reposts Media

Dongyeop Kang (DK)
@dongyeopkang
11h
How much time do you struggle to edit AI-generated "slop" in your teams’ work? We built a system to help automate the process. Our Scientific Slop Index (SSI) measures scientific slop in a draft not only looking at words but also investigating their scientific reasoning process:
Yerim Oh
@OhYerim24061
Oct 4
AI detectors read your sentences. We read your scientific reasoning and score its slop. Every ICLR year, 2017–2025: papers with more our "scientific slop" got lower reviewer ratings and were rejected more often. Calculate yours: yerimoh.github.io/scientific-slo…
Dongyeop Kang (DK)
@dongyeopkang
Jul 6
Not attending ACL this year but many of @MinnesotaNlp members presenting cool papers on LLM evaluation, ! Highly recommend talking to them!! Young-Jun Lee (@passing2961): AI4Science, evolution scaffold Julia Christenson (julia-christenson.github.io): LLM's effects on essay writing
1
Dongyeop Kang (DK)
@dongyeopkang
Jul 6
If you’re at ACL 2026 today, stop by @elihc_2706's poster presentation on "ScholaWrite." July 6, Poster Session E, 11:00 AM to 12:30 PM Paper: aclanthology.org/2026.acl-long.… Collecting human process data from writing (or other other complex workflows) can help train more cognitively
Khanh Chi Le (Chi)
@elihc_2706
Jul 4
Do LLMs actually write like scientists? ScholaWrite captured 62K real keystrokes across 5 papers, exposing how scholarly writing really unfolds — and showing LLMs still can't keep up with it. 📍 #ACL2026, Mon 7/6, Poster Session E 📄 minnesotanlp.github.io/scholawrite 🧵
Dongyeop Kang (DK)
@dongyeopkang
Jul 1
Check out Young-Jun (@passing2961)'s first work at @MinnesotaNlp , showing the cross-domain transferability of optimization tasks. "Can experience from winning IMO help you win IPO or IOI?" Trajectory learning as midtraining may be a promising direction for bridging the
Young-Jun Lee
@passing2961
Jul 1
🤔Can experience optimizing in one domain (e.g., faster kernel design) transfer to a very different domain (e.g., open mathematical conjectures)? 💡We found that this kind of cross-domain optimization is possible through Evolution Fine-Tuning (EFT). 🤗huggingface.co/papers/2606.29…
GIF
Dongyeop Kang (DK)
@dongyeopkang
Jun 4
We’re looking for professionals who use AI in their workplace to participate in our study. Please contact @shirley_ah_ and help spread the word!
Shirley A. Hayati ✈️ ACL
@shirley_ah_
Jun 4
📢 Calling Professionals! We're studying how domain experts (e.g., healthcare professionals, managers, designers) use AI agents in their work. Participants who complete the study will receive $75 gift card. Please fill out our participant interest form! forms.gle/J8g4Gnwc4RZaz8…