Joseph Jeesung Suh

🎉 Thrilled to have two papers accepted to ACL 2026 main! 1. Graph-based models match LLMs on close-ended human simulation tasks with far less compute & greater transparency 2. (oral) How to allocate human samples towards fine-tuning vs post-hoc rectification in simulation

🎉 Excited to share that GEMS is accepted to ACL 2026 main! We show that a lightweight GNN can match or outperform LLMs at simulating human behavior in discrete-choice settings — with multiple advantages, including efficiency and transparency. Paper: arxiv.org/abs/2511.02135

@profjoeyg @woosuk_k @vllm_project Congratulations Woosuk! Wishing you all the best in your future journey

I am excited to attend @woosuk_k's thesis defense on PagedAttention and the @vllm_project. Congratulations on such an incredibly high impact project!

📢 Come work with me at UC Berkeley @berkeley_ai! I’m recruiting PhD students in @Berkeley_EECS and @UCJointCPH. I work on AI for social good, simulating humans with AI, human-AI interaction, and applications in public health & social science. serinachang5.github.io

(11/11) For people who are interested, here is a link: Paper: arxiv.org/abs/2511.02135 Github: github.com/schang-lab/gems Huge thanks to my amazing PI @serinachang5 and collaborator @SuhongMoon.

(10/11) Takeaway 🥡 If your simulation task is a discrete choice with relational structure, try GEMS 💎 before spinning up a 70B param model. You might get similar (or better!) accuracy with a fraction of the compute and better debug-gability!

LLMs have dominated recent work on simulating human behaviors. But do you really need them? In discrete‑choice settings, our answer is: not necessarily. A lightweight graph neural network (GNN) can match or beat strong LLM-based methods. Paper: arxiv.org/abs/2511.02135 🧵👇

🤔 Do LLMs exhibit in-group↔out-group perceptions like us? ❓ Can they serve as faithful virtual subjects of human political partisans? Excited to share our paper on taking LLM virtual personas to the *next level* of depth! 🔗 arxiv.org/abs/2504.11673 🧵

💡New preprint & Python package: We use sparse autoencoders to generate hypotheses from large text datasets. Our method, HypotheSAEs, produces interpretable text features that predict a target variable, e.g. features in news headlines that predict engagement. 🧵1/

New Paper: We unlock AI Evaluation with explanatory and predictive power through general ability scales! -Explains what common benchmarks really measure -Extracts explainable ability profiles of AI systems -Predicts performance for new task instances, in & out-of-distribution 🧵

@Ugo_alves Hi! The largest model we fine-tuned was 70B, Llama-3-70B

@JosephJSSuh Is my understanding correct that the largest model you’ve fine tuned was 13b?

Can LLMs assist public opinion survey designs by predicting responses? We fine-tune LLMs on our new large-scale survey response dataset, SubPOP, which reduces the distributional gap between human-LLM predictions by up to 46% 📊 A 🧵 on our findings: 👇