Ritwik Gupta 🇺🇦
@Ritwik_G
Followers
3K
Following
8K
Media
209
Statuses
2K
Postdoc @berkeley_ai | Incoming Assistant Professor @umdcs | Technical Director @DIU_x
Berkeley, CA
Joined June 2012
I am recruiting Ph.D. students at @umdcs starting Fall 2026! I am looking for students in three broad areas: (1) Physics-integrated computer vision (2) VLMs with constraints (2) Dual-use AI policy We're ranked #3 in AI on @CSrankings! Specific details in 🧵
11
106
394
Deadline is December 5! Link: https://t.co/E3f3rXVqDd Fee waiver info: 1. https://t.co/PKg7pElX10 2. https://t.co/ChpuxwNDQO (for domestic students; Nov. 15 deadline)
btaa.org
0
1
7
Policy topics I'm exploring: 1. Open source analysis of AI advancements 2. Civil-military fusion and bounding the meaning of "dual-use" in AI 3. Alternatives to hardware export controls
1
0
4
VLM topics I'm exploring: 1. Guaranteed generation given (learned) constraints 2. Skill synthesis/distillation
1
0
4
CV topics I'm exploring: 1. Learning directly from phase history/active imaging echoes such as radar and sonar (students with a background in signals/EE and CV wanted!) 2. VLMs with inductive physical biases 3. Multi-sensor fusion (e.g. EO / radar / thermal)
1
0
3
Open Rank Faculty Positions: @UMDCS We invite applications for Assistant, Associate & Full Professors across areas ranging from cybersecurity to data science and computer systems. 🗓️ Best consideration: Dec. 31, 2025 Apply now: https://t.co/4DXluYgaFz
1
3
5
Wen-Han is interested in grad school and will be an absolute pull for anyone looking to do work in the space of multimodal reasoning! Certainly I will be trying to recruit him :)
AI can now see, reason, and segment the Earth. 🌍 Meet LISAt, our #NeurIPS2025 Datasets & Benchmarks paper - the first foundation model that turns language queries into pixel-level satellite segmentations. 🛰️ (1/n) 🔗 https://t.co/ApVZgGF0cU
@NeurIPSConf @berkeley_ai
1
2
31
Proud advisor moment 😊 Congrats @Songwei_Ge for winning the Larry S. Davis Doctoral Dissertation Award @umdcs! Songwei is now cooking as a research scientist at @reve. Looking forward to amazing work!
12
3
91
Humans handle dynamic situations easily, what about models? Turns out, they break in three distinct ways: ⛔ Force Stop → Reasoning leakage (won’t stop) ⚡️ Speedup → Panic (rushed answers) ❓ Info Updates → Self-doubt (reject updates) 👉Check out https://t.co/wKrnsMkiFY
5
21
68
✨Introducing ECHO, the newest in-the-wild image generation benchmark! You’ve seen new image models and new use cases discussed on social media, but old benchmarks don’t test them! We distilled this qualitative discussion into a structured benchmark. 🔗 https://t.co/wJmmEY8TFQ
3
31
114
I’m going to be at BayLearn today. If you are there, hit me up!
0
0
2
I’m at #COLM2025 🇨🇦presenting “Hidden in Plain Sight: VLMs overlook their vision representations” as a Poster and Oral! Also honored to win Outstanding Paper here and Best Paper @ CVPR EVAL-FoMo 2! Come chat at poster 12 (Wed AM) about building perceptual representations! (1/3)
3
11
61
1/ Can molecular AI move past hard-coded Graph Neural Networks and embrace scalable Transformers that discover molecular structure on their own? We demonstrate that you can train a 1B parameter Transformer model without any graph priors or physical inductive biases. And
6
32
172
Thrilled that HRC Faculty Director @KAlexaKoenig is one of the inaugural members of Women Transforming Global Security, a new feminist global network addressing crises related to nuclear weapons proliferation, climate change, human rights, and more:
transformingsecurity.org
Women Transforming Global Security represents a groundbreaking alliance of distinguished women leaders pioneering a comprehensive framework for addressing
2
3
5
Life update: before I start at UMD, I'm spending a year doing a post-doc with the amazing @ask1729 at Berkeley! I'm working on ML for chemistry: ML force fields, spectral embeddings, and more! So much to learn, so much to do, and so excited to join this amazing team.
1
1
28
REOrder was accepted to @NeurIPSConf! Images are getting bigger, and long sequence vision transformers are becoming more common. We show that ordering of input patches greatly affects performance (10-13% accuracy!) and introduce REOrder to find an optimal ordering. @kutsch_d
Ever wondered if the way we feed image patches to vision models is the best way? The standard row-by-row scan isn't always optimal! Modern long-sequence transformers can be surprisingly sensitive to patch order. We developed REOrder to find better, task-specific patch sequences.
0
2
14
Great takes by Sreejan, as always
People ask me what my take on AI is these days, so here's a thread on a new idea I dont think anyone who's working in AI is really thinking about. 1/12
1
0
2
I spent an amazing week in beautiful (and stormy) Taipei speaking about AI-military integration and perception and reasoning in chaotic environments. Thank you to the Taiwan authorities, @NTU_TW, and @AcadSinica for hosting me!
1
0
10