Ismini Lourentzou
@Ismini_L
Followers
326
Following
364
Media
25
Statuses
225
Assistant Prof. @iSchoolUI Multimodal ML, Computer Vision, NLP Research. Previously @VT_CS @SanghaniCtrVT @IBMResearch. PhD @illinoisCDS. Opinions are my own.
Joined September 2013
๐ฟ๐ง
Former Google engineer David Budden claims to have solved one of humanity's most difficult mysteries and is having it verified not by humans, but by a computer: The Navier Stokes Millennium Problem. Should his proof prove to be correct, it would have a significant impact on
0
0
0
Thank you to the @Google Tunix team for the collaboration and the @googlecloud TPU support behind this work! https://t.co/6vY8NuRuxa fDPO from our #NeurIPS2025 SpatialReasoner-R1 paper is open-sourced in Tunix ๐ #JAX #OpenSource
opensource.googleblog.com
Researchers from UIUC and Google introduce SpatialReasoner, a VLM with enhanced spatial reasoning. They used Tunix on TPUs to develop fDPO.
0
1
3
Excited to share our #AAAI2026 Oral paper Hierarchical Dataset Selection for High-Quality Data Sharing! ๐ค How do you decide which datasets to train on when data comes from many noisy, heterogeneous sources? ๐งต
1
0
4
Whoa! Love this wormhole experience of teleporting between Marble worlds! ๐คฉ
Exploring portals through marble generated scenes, powered by Spark and a bit of camera shader magic.
37
65
627
Thank you to everyone who stopped by our @NeurIPSConf poster on SpatialReasoner! Lots of exciting things cooking in PLAN Lab - stay tuned!๐And if you missed our poster, hereโs our paperโฆ as a groovy song ๐ถ https://t.co/7EEeer2WMt ๐ง๐ธ
0
0
3
NeurIPS 2025's top 50 paper contributors show China and the US neck-and-neck. In the US, corporate labsโGoogle DeepMind, Meta, Microsoftโnow rival top schools like Stanford, CMU, MIT, showing a major shift to industry-led AI research. Chinaโs leaders remain academic: Tsinghua,
30
336
2K
Happening now! @NeurIPSConf Come learn about spatial reasoning VLMs
0
0
3
Heading to @NeurIPS2025 ๐ฉ๐ฝโ๐ป looking forward to catching up with friends and colleagues. @SivanShen will be presenting our spatial reasoning work ๐ Exhibit Hall C/D/E โ Poster #4812 ๐๏ธ Wed, Dec 3 | 11amโ2pm PST ๐ https://t.co/4YK9jJi6iM If youโre interested in VLMs, reasoning,
0
1
4
๐ https://t.co/82zprHAqs6 โฐ Submissions Due Dec 16, 2025 ๐๏ธ Join us in March 2026 @ WACV'26 in Tucson, AZ!
0
0
0
๐ LLVM-AD @ WACV 2026 CFP is live! ๐ Working on autonomous driving? The LLVM-AD workshop invites papers on LLMs/VLMs for autonomous driving, including multimodal planning, scene understanding, language-driven simulation, datasets, and humanโautonomy teaming. ๐๐๐ผ๐๐พ
1
0
0
Excited to finally share our #NeurIPS2025 paper "๐ฎPurpCode: Reasoning for Safer Code Generation"!ย ๐ ๐ First post-training recipe for training safe code reasoning models ๐ SOTA for cybersafety + utility, outperforming Sonnet 4, o4-mini, R1 ๐ฅ Winner of 2025 Amazon Nova AI
1
6
17
Very proud of my student @SivanShen for his first-author #NeurIPS2025 SpatialReasoner-R1 paper ๐ An amazing start to his #siebelschool PhD! ๐Huge thank you to our co-authors Yuanzhe Liu, Jingyuan Zhu, @IrohXu, Xiaofeng Zhang, Yixiao He, @wenmingye, @RehgJim
๐Can we tackle spatial queries just from 2D images? Yes, thanks to LongCoT reasoning!๐บ๏ธIntroducing ๐๐ฉ๐๐ญ๐ข๐๐ฅ๐๐๐๐ฌ๐จ๐ง๐๐ซ-๐๐, a vision-language reasoning model that sets new SoTA on SpatialRGPT-Bench. ๐ https://t.co/0sWj0gg1MO ๐ https://t.co/NFeMLiU1qm
0
1
9
2 days left (September 19, 2025 AoE) to submit Resources (Data/Software); JCDL 2025 invites submissions to a dedicated Resources Track, focused on public datasets, software & benchmarks. CFP: https://t.co/ZWY8iLps8N
#JCDL2025 #digitallibraries
0
3
2
Call for Resources (Data/Software); an interesting and exciting CFP and venue especially for people working in the AI4Science. Do check it out !!
2 days left (September 19, 2025 AoE) to submit Resources (Data/Software); JCDL 2025 invites submissions to a dedicated Resources Track, focused on public datasets, software & benchmarks. CFP: https://t.co/ZWY8iLps8N
#JCDL2025 #digitallibraries
0
1
0
Congratulations to #iSchoolUI Assistant Professor @Ismini_L and everyone on the PurpCorn-PLAN team on this major accomplishment! ๐
#ILLINOIS PurpCorn-PLAN has won the @amazon Nova AI Challenge, claiming an impressive $250,000 prize! ๐ถ๐๐ท This team of 11 PhD & MS students from @uofigrainger Siebel School & @iSchoolUI is led by profs Gang Wang, Lingming Zhang & Ismini Lourentzou.
0
2
2
#ILLINOIS PurpCorn-PLAN has won the @amazon Nova AI Challenge, claiming an impressive $250,000 prize! ๐ถ๐๐ท This team of 11 PhD & MS students from @uofigrainger Siebel School & @iSchoolUI is led by profs Gang Wang, Lingming Zhang & Ismini Lourentzou.
amazon.science
University teams battle to harden and hack AI coding assistants in head-to-head tournamentย
0
4
11
๐ ๏ธ What's new? โ๏ธFine-grained Direct Preference Optimization (fDPO) for spatial reasoning โ
Multi-Model MCTS for generating high-quality LongCoT spatial reasoning data ๐In collaboration with students and colleagues from UPenn, Shanghai Jiao Tong University, UIUC, and Google.
0
0
0
๐Can we tackle spatial queries just from 2D images? Yes, thanks to LongCoT reasoning!๐บ๏ธIntroducing ๐๐ฉ๐๐ญ๐ข๐๐ฅ๐๐๐๐ฌ๐จ๐ง๐๐ซ-๐๐, a vision-language reasoning model that sets new SoTA on SpatialRGPT-Bench. ๐ https://t.co/0sWj0gg1MO ๐ https://t.co/NFeMLiU1qm
1
0
5