Fanghui Liu (@fanghui_sgra) 's Twitter Profile
Fanghui Liu

@fanghui_sgra

Assistant Professor @WarwickDCS🇬🇧 | prev. @EPFL🇨🇭@KU_Leuven🇧🇪 @sjtu1896🇨🇳| kernel, learning theory| co-organizer Foundation of AI seminar @FAIS_Warwick

ID: 1182893719332282368

linkhttp://www.lfhsgre.org calendar_today12-10-2019 05:40:27

100 Tweet

267 Followers

235 Following

Fanghui Liu (@fanghui_sgra) 's Twitter Profile Photo

#WarwickFoundationsAI Thanks for Bruno Bruno Loureiro from CNRS/ENS about his talk “Feature Learning and computational limitations in learning with multi-index models” today! FoundationsAISeminars

Fanghui Liu (@fanghui_sgra) 's Twitter Profile Photo

I will give the presentation today 4pm at #ICML2025 Oral session: Learning dynamics 2 @ West Ballroom B! Here is the poster and long-version slides (lfhsgre.org/files/talk_LoR…) if you’re interested in.

I will give the presentation today 4pm at #ICML2025 Oral session: Learning dynamics 2 @ West Ballroom B!

Here is the poster and long-version slides (lfhsgre.org/files/talk_LoR…) if you’re interested in.
Mikhail Samin (@mihonarium) 's Twitter Profile Photo

🚨 According to a friend, the IMO asked AI companies not to steal the spotlight from kids and to wait a week after the closing ceremony to announce results. OpenAI announced the results BEFORE the closing ceremony. According to a Coordinator on Problem 6, the one problem OpenAI

🚨 According to a friend, the IMO asked AI companies not to steal the spotlight from kids and to wait a week after the closing ceremony to announce results. OpenAI announced the results BEFORE the closing ceremony.

According to a Coordinator on Problem 6, the one problem OpenAI
Fanghui Liu (@fanghui_sgra) 's Twitter Profile Photo

Surprisingly, even free-tier ChatGPT Vision can tackle some easy-level IMO 2025 problems—with just light human supervision or verification. 🔍🧠 Try it yourself: chatgpt.com/share/687fdfb9…

Surprisingly, even free-tier ChatGPT Vision can tackle some easy-level IMO 2025 problems—with just light human supervision or verification.
🔍🧠
Try it yourself: chatgpt.com/share/687fdfb9…
Fanghui Liu (@fanghui_sgra) 's Twitter Profile Photo

We had one work published on ICML’25 (oral) demonstrating that one-step gradient from full fine-tuning is sufficient for LoRA for some cases! arxiv.org/abs/2502.01235 With my PhD student Yuanhe Zhang and collaboration with Yudong from Madison.

Fanghui Liu (@fanghui_sgra) 's Twitter Profile Photo

🎓 PhD positions in Foundations of AI & ML 📍 University of Warwick, CS Dept (FAM Division) Department of Computer Science - Warwick 🗓️ Apply by 14 Nov 2025 👉 lnkd.in/e75-MPZc Join one of the UK’s top 4 CS departments (REF 2021)! #PhD #AI #MachineLearning #AcademicJobs #Warwick

Fanghui Liu (@fanghui_sgra) 's Twitter Profile Photo

🚀 We’re organizing @IJCV special issue on “Post-Training in LLMs for Computer Vision”! With guest editors Grigoris Chrysos Rene Vidal Dacheng Tao (NTU) and Philip Torr (Oxford) 📘 Learn more & submit your work: ijcvsi25pt.github.io 📅 Submission ddl: July 15, 2026

🚀 We’re organizing @IJCV special issue on “Post-Training in LLMs for Computer Vision”! With guest editors <a href="/Grigoris_c/">Grigoris Chrysos</a> <a href="/vidal_rene/">Rene Vidal</a> Dacheng Tao (NTU) and Philip Torr (Oxford)

📘 Learn more &amp; submit your work: ijcvsi25pt.github.io

📅 Submission ddl: July 15, 2026
Fanghui Liu (@fanghui_sgra) 's Twitter Profile Photo

Check out Yuanhe Zhang 's thread on our recent work exploring a new pipeline to model step-level reasoning, a “Goldilocks principle” that balances free-form CoT and formal systems! 👇🏾

Check out <a href="/yuanhezhang6/">Yuanhe Zhang</a> 's thread on our recent work exploring a new pipeline to model step-level reasoning, a “Goldilocks principle” that balances free-form CoT and formal systems! 👇🏾