Hongli Zhan

@HongliZhan

🤖Shaping model behaviors @IFM_AI, Silicon Valley Lab // 🤘Ph.D. @UTAustin

San Francisco Bay Area

honglizhan.github.io

Joined May 2020

532 Following

613 Followers

Replies Reposts Media

Pinned
Hongli Zhan
@HongliZhan
Apr 3
PhD defended at UT Austin today.🤘 The best thing was having an advisor who believed in me before I believed in myself. Jessy taught me how to write, how to think, and how to chase research ideas. Then the rest followed. Thank you, @jessyjli
11
Hongli Zhan
@HongliZhan
Sep 3
Lots of frontier models dropping lately, but how many let you see how they were built? K2-Horizon releases training data, recipes, code, checkpoints for all 6 sizes, 0.9B dense to 375B MoE. Not just open weights, open everything. Really proud of this team! 🙌 👇 Go try them!
Institute of Foundation Models
@IFM_AI
Sep 3
Introducing K2 Horizon: a connected fleet of six foundation models ranging from 0.9 billion to 375 billion parameters. - Frontier performance: Across coding and agentic tasks, K2 Horizon delivers top-tier performance in every size class—with the 0.9B, 3.7B and 7B models setting
1
Hongli Zhan
@HongliZhan
Aug 11
Excited to be part of the team. And what a way to start 🫡
Institute of Foundation Models
@IFM_AI
Aug 10
A new space for the work ahead. We’re excited to share a look at IFM’s new office in Sunnyvale, a place for our team to collaborate, exchange ideas, and continue advancing open foundation AI. This new space marks the next stage of IFM’s growth as we expand our global team and
Hongli Zhan
@HongliZhan
Jun 9
Fun collab with @Gz_The_Divine! Hallucination detectors say they evaluate reasoning, but many are just reading the final answer. Swap or remove it, keep the reasoning intact, and the scores break. What if we look at the reasoning trajectory instead? 🔍
GZ The Georgia Tech Basketball Fan
@Gz_The_Divine
Jun 9
🚨New paper! Your hallucination detector says it evaluates reasoning. But what if it's just peeking at the final answer? We tested this: keep the reasoning, only change the answer. Many detectors' scores shift dramatically 🧵
Hongli Zhan
@HongliZhan
May 3
1k+ downloads each on the MINT empathy models since release 🔥 Encouraging to see the interest in our work! tl;dr: In multi-turn empathic dialogue, LLMs reuse the same discourse moves far more often than humans do; MINT uses RL to diversify them. Give it a try!👇
Hongli Zhan
@HongliZhan
Apr 29
New paper! 🏁 My final one from my PhD at UT Austin. 🦜LLMs sound empathic, but they keep saying the same thing over and over. Not just the same words, the same discourse moves, turn after turn. We found that LLMs repeat the same discourse moves at nearly 2x the rate of human
1

Hongli Zhan

Hongli Zhan’s posts