Thomas De Min

🎉[openings] I’m hiring postdoctoral researchers to join our @FunAILab at UTN through the Alexander von Humboldt Research Fellowship (@AvHStiftung), via the Henriette Herz Scouting Programme. As a Henriette Herz Scout, I can nominate outstanding international researchers for this fellowship route. I’m especially keen to hear from candidates working on multimodal learning, video and image pretraining, and post-training. Fellows would be hosted in our lab at UTN and work closely with us on these topics. Key requirements: * finished your doctoral studies less than 4 years ago or will finish in the next 6 months * did not live/work in Germany in the last 10 years * applications from female, trans* and/or non-binary candidates are highly encouraged! Interested? Please send a short note with your CV, PhD year, current affiliation, 2–3 key publications, and a few lines on how your work connects. Please share! 🔀

There were bugs in DeepSpeed and OpenRLHF that reduce SFT performance, and this affected several studies.

This is a good guy... Hire him if you can! 😏

Huge congrats to our colleague and friend @lucazanella95 on his graduation! 🎓 Brilliant researcher, great teammate, and a wonderful person, we’re proud to celebrate this milestone with you, Luca! Looking forward to what comes next 💪

We are so back 😎

👇👇👇

Current VLMs lack fine-grained spatial perception. Introducing TerraScope [CVPR 2026]: a pixel-grounded reasoning framework for Earth Observation that interleaves segmentation masks with chain-of-thought for precise geospatial analysis. 🌐 shuyansy.github.io/terrascope/ #CVPR2026 #VLM

5/5 And finally, a huge thanks to my amazing co-authors Subhankar Roy (@sroy907), Stéphane Lathuilière (@Steph_lat), Elisa Ricci (@eliricci_), and Massimiliano Mancini! Paper: arxiv.org/abs/2603.19466 Code: github.com/tdemin16/proac… Data🤗: tdemin16/ProactiveBench

4/5 The good news: proactiveness can be learned. RL post-training with GRPO yields strong improvements and generalizes to unseen scenarios, even for the worst-performing baseline.

Can MLLMs ask for help? 🤔 When facing an unanswerable visual query, models typically hallucinate or abstain. We argue there's a better option: being proactive and requesting a simple user intervention. Introducing ProactiveBench! 🧵

Auch... You just missed #ECCV2026 deadline!! 🤷 Don't worry, you still have a few days left to submit your work to the MULA workshop at #CVPR2026!! See you in Denver! 🗓 Deadline: March 9 🔗 Details: mula-workshop.github.io

🚨 One week to go! 🚨 Don’t miss your chance to submit a novel contribution to the MULA workshop at #CVPR2026. 🗓 Deadline: March 9 🔗 Details: mula-workshop.github.io

@eccvconf Hi, can we use wrapfigure in the paper? Or, more generally, can we wrap text around figures and tables?

For all #ECCV2026 deadlines, check out our webpage! Deadlines: eccv.ecva.net/Conferences/20…

CVPR decisions are out, so congrats to everyone with an acceptance! 🎉 If your paper didn’t make the main conference but you believe in the idea and want to discuss it with the community, consider submitting to the MULA Workshop!! Info here: mula-workshop.github.io #CVPR2026

@eccvconf Silly question, can we use the wrapfig package to wrap text around figures?

What’s New at #ECCV2026 Malmo 🇸🇪? Please read the important policy updates (especially with regard to ECCV 2024) on our “What’s new?” page. Notably, this year, we introduced “Contribution Types”, a mechanism for tagging submissions &reviewers to facilitate fair evaluation. 1/2

Here is MULA 2026 great speakers: • Hedvig Kjellström • Georgia Gkioxari (@georgiagkioxari) • Serena Yeung-Levy (@yeung_levy) • Yuki Asano (@y_m_asano) • Ranjay Krishna (@RanjayKrishna) • Marcella Cornia (@marcella_cornia) More information ➡️ shorturl.at/rSZdP @CVPR

🚀 Call for Papers! The 9th Multimodal Learning and Applications Workshop (MULA 2026) is coming to @cvpr. If you’re working on multimodal learning, this is your stage! 🗓️ Submission deadline: March 9, 2026 (AoE) More information ➡️ mula-workshop.github.io

Group-robust Machine Unlearning Thomas De Min, Subhankar Roy, Stéphane Lathuilière, Elisa Ricci, Massimiliano Mancini. Action editor: Vidya Muthukumar. openreview.net/forum?id=StSq7… #robust_machine_unlearning #unlearning #unlearn