Tae Soo Kim

Improving LLM outputs is difficult because evaluating these outputs is difficult... Introducing EvalLM ⚗️ - an interactive system for evaluating LLM outputs on user-defined and application-specific criteria 📏 🔗: evallm.kixlab.org 📄: arxiv.org/abs/2309.13633 (1/n)

I'm in Yokohama for #CHI2025 to present my work #Proxona! 😊 I'll be presenting at 9:12 AM in Room G314+G315. If you're interested in user simulation & prototyping, data-driven personas, conversational interactions, or supporting sensemaking and ideation, please come!

How should we update AI memory of user intent as intent evolves? 🤖💭 How can we auto-update memory docs like Cursor rules as a user interacts with an AI agent over time? Check out our pre-print: Semantic Commit: Helping users update intent specifications for AI memory at scale!

Huge congratulations to @tilekbayb on successfully defending his M.S. thesis! Bekzat will be joining KIXLAB as a Ph.D. student this Fall!

Please RT! We're seeking data analysts using AI in their workflows. We would love to observe your process and learn what has been exciting or challenging. Your insights will help us build better tools. If you are willing to sit down with us (on zoom) for a couple of hours…

AI should not replace browsing, but scale it. #Orca turns the web into your canvas and personal workspace. Work across dozens of pages, delegate to AI agents by your side, and synthesize on the fly. Welcome to 𝗕𝗿𝗼𝘄𝘀𝗶𝗻𝗴 𝗮𝘁 𝗦𝗰𝗮𝗹𝗲 hci.ucsd.edu/orca w/@HaijunXia

안녕하세요, @kixlab_kaist 에서 만든 📢 공약쏙쏙 시스템📢을 소개합니다! (promisescope.com) 🤔공약쏙쏙? #대선2025 공약을 바탕으로 공약이 가져올 수 있는 사회적 영향, 그리고 '나'에게 어떤 변화가 생길 수 있을지 LLM (GPT-4o) 을 활용해 자동으로 보여주는 시스템입니다. (1/3)

🆕paper: LLMs Get Lost in Multi-Turn Conversation In real life, people don’t speak in perfect prompts. So we simulate multi-turn conversations — less lab-like, more like real use. We find that LLMs get lost in conversation. 👀What does that mean? 🧵1/N 📄arxiv.org/abs/2505.06120

🤖LLMs are being integrated everywhere, but how do we ensure they're delivering meaningful user experiences? In our #chi2025 paper, we empower designers to think about this via 🎨designerly adaptation🎨 of LLMs and built a Figma widget to help! 📜 arxiv.org/abs/2401.09051 🧵👇

i’m at #chi2025 and i’ll be on the industry job market later this year! i work in human-ai interaction. my prev projects focused on design tools. i love design. i love user interfaces. i trained myself to become an ai engineer to push our tools further. i believe ai is on…

In this era of vibe coding fueled by LLMs, our #CHI2025 paper, led by amazing @StevenHe918, asked a simple question: Can we really get away with *not* using gold labels? Paper: arxiv.org/abs/2502.11267 1/n

I've been quiet for a while on the social media front -- it's been a very busy two years with an international move and setting up a new lab at KAIST, but I wanted to take a moment to highlight some of my students' fantastic new work in the domain of online safety.

🤖👩🏻‍💻Proactive AI tools like @GitHubCopilot @cursor_ai @allhands_ai promise to assist developers by anticipating their needs and automating engineering processes—but do they truly help? We evaluated three design probes to explore the trade-offs of proactive AI programming support

🌸New #CHI2025 paper alert!! 🤖👩‍🏫 How can teachers ensure that AI tutors work for every student? Our team envisions a future where teachers build chatbots (e.g., AI tutors) tailored to their curriculum and students. 📄 Paper: arxiv.org/abs/2410.04078

Now that we got a Best Paper award at #CHI2025 🏆, I have to tweet about this! We studied how 39 software professionals - designers, PMs, and engineers - prototype GenAI applications. (I had fun co-conspiring with Divy @divy93t on this paper.) Paper: arxiv.org/abs/2402.17721 1/n

When learning new skills, people often watch *multiple* how-to videos to understand different approaches and details. In our #IUI2025 paper, we introduce VideoMix — a system that uses vision-language models to aggregate videos into one organized, easy-to-digest experience.

Midjourney announces Modifying Large Language Model Post-Training for Diverse Creative Writing

#CHI2025 paper from Midjourney Storytelling Lab 🫱🧸 🦖🫲What if we can tell a story by playing with toys? 🫱🧸 🦖🤖What if we can collaborate with AI on toy-playing+storytelling? Toyteller enables toy-playing-based storytelling with AI. Learn more: mj-storytelling.github.io/project/toytel… (1/9)

Photoshop for text. In our #CHI2025 paper “Textoshop”, we explore how interactions inspired by drawing software can help edit text. We consider words as pixels, sentences as regions, and tones as colours. #HCI #NLProc #LLMs #AI Thread 🧵 (1/10)

Malleable Overview-Detail Interfaces at #CHI2025 How do we design malleable interfaces in the age of GenAI? How can we pull information instead of navigating to find it? How can we create our own abstractions of information? Check out our preprint: arxiv.org/abs/2503.07782

Exciting news 🪩 I'll present our upcoming paper Proxona: Supporting Creators' Sensemaking and Ideation with LLM-Powered Audience Personas at #CHI2025 in Yokohama! 🇯🇵 Can we simulate virtual yet plausible and meaningful audience personas for content creators?