Bruno Ribeiro (@brunofmr) header
Bruno Ribeiro (@brunofmr) profile picture

Associate Professor of Computer Science @Purdue, interested in graphs, invariances, causality, and OOD robustness in ML

West Lafayette, IN
Joined December 2009
Photo by Bruno Ribeiro (@brunofmr)Photo by Bruno Ribeiro (@brunofmr)Photo by Bruno Ribeiro (@brunofmr)Photo by Bruno Ribeiro (@brunofmr)Photo by Bruno Ribeiro (@brunofmr)Photo by Bruno Ribeiro (@brunofmr)Photo by Bruno Ribeiro (@brunofmr)Photo by Bruno Ribeiro (@brunofmr)Photo by Bruno Ribeiro (@brunofmr)Photo by Bruno Ribeiro (@brunofmr)
🚨Tutorial on Positional and Structural Graph Representations invariances.org/graph-repres… Given at the National Institute for Theory and Mathematics in Biology. Thanks to @ClaireDonnat and Olga Klopp for the invitation! @HaggaiMaron, the tutorial is finally out!
5
24
3,034
Bruno Ribeiro retweeted
We’re hiring postdocs at @StanfordAILab! Join the Leskovec Lab, in collaboration with the Good Lab, to build foundation models and AI agents for biology. Help turn multimodal biomedical data into scientific discoveries. Apply: snap.stanford.edu/apply/inde…
4
41
2
271
28,683
Bruno Ribeiro retweeted
the full geometric deep learning book draft is now complete. all chapters available online on our website (link in thread). 🫡 @mmbronstein @TacoCohen
22
141
18
1,031
55,158
The more AI thinks, the more convinced it is it knows the answer. Maybe a failure in reward training? Long thinking chains associated with positive rewards, so the model has learned to treat chain length as a proxy for finding the answer? arxiv.org/abs/2610.28340
1
1
256
In fairness, people may have the same bias: the more they talk, the more convinced they are that they're right. But that's possibly spurious too: More thoughtful people just talk less.
55
After using & building AI systems for personal use, it feels like a weird mix of an odd PhD student, secretary, and politician: It knows a lot, but makes odd mistakes Handles any tedious work fine Resists gentle corrections, but correct it hard and swings completely the other way
3
224
#ICLR2027 authors: if you have solid evidence that a reviewer is nitpicking, flag it to your AC. In rebuttal, be specific: quote the exact claim, then rebut it: "This is standard in my area, see [refs], pp. 12–14." "Feels AI-written" or "here is an AI score" is not enough.
6
8
184
22,240
We will be looking for nitpicking. However, while AI for reviewing has many shortcomings, a broken clock is right twice a day. E.g., AI flags that a table bolds the best performance numbers with no confidence intervals. Yes, that's a real problem. A rebuttal needs substance.
7
2,373
Bruno Ribeiro retweeted
I’m very excited to share @nvidia Kumo Tabular, a new family of foundation models for tabular data. Kumo Tabular establishes the new Pareto frontier across the entire accuracy–inference-time tradeoff. Just as importantly, we are releasing it openly: open weights, open-source software, and a permissive license for commercial use. HuggingFace: huggingface.co/nvidia/Kumo-T… GitHub: github.com/NVIDIA/structured…
31
154
34
1,110
129,350
On September 6 we celebrated Judea Pearl's 90th birthday at UCLA, and the morning session is now online. It was a day to say thank you to Judea, and an honor to organize it with Rina Dechter and Hector Geffner. Students, collaborators, and friends came from across the world, and the affection in the room was the real subject of the day. The morning had a keynote by @tdietterich, talks by Chris Hitchcock, @DaphneKoller, and me, a panel with @DanRothNLP, @erichorvitz , and @pmddomingos, remarks by Chancellor @UCLAchancellor, and personal reflections from students, collaborators, and friends. youtube.com/watch?v=dSn6wsUF… Thank you to Zvi Galil, Leonard Kleinrock, Deborah Estrin, Moises Goldszmidt, Kaoru Mulvihill, Ayala Or-El, and Jason Cong, who spoke in the room, and to Moshe Vardi, Stuart Russell, Uri Sivan, David Harel, vint cerf, Warshel Arieh, Bernhard Schölkopf, Steven Sloman, guido imbens, and Dan Geiger, who sent messages, and to everyone who came. The afternoon session will follow.
1
9
2
60
24,603
Bruno Ribeiro retweeted
I'm hiring a postdoc! The position is funded by my new ERC Starting Grant, GDLNeuralArt, and based in my group at the Technion. We work on learning on weight spaces, internal representations and gradient spaces: treating neural networks themselves and their artifacts as data. 🧵
4
21
2
98
7,425
I would NEVER provide feedback on a codex or claude session, regardless of privacy settings. Feedback = explicit permission for data use
🤣 The 2026 Trojan Horse🐴 runs on a terminal 🖥️ 1. If you don't control the inference, the idea isn't yours. 2. Re-read "Privacy policy" knowing that, say, "purely AI-generated material is not copyrightable"; now adversarially define "yours", "train", and "pure" against you 🫠
253
🤣 The 2026 Trojan Horse🐴 runs on a terminal 🖥️ 1. If you don't control the inference, the idea isn't yours. 2. Re-read "Privacy policy" knowing that, say, "purely AI-generated material is not copyrightable"; now adversarially define "yours", "train", and "pure" against you 🫠
We’re giving scientists, mathematicians, and engineers free access to our frontier models—starting with 10,000 researchers and expanding to 100,000 through 2027. ChatGPT for Academic Researchers is built to accelerate discovery across disciplines.
1
1
1
5
1,579
Now in a more serious note, my heart really goes out to the folks that used AI cloud credits for innovative research. Hopefully universities will start better advising researchers regarding AI use and IP.
95
Congrats to all authors. And special congrats to @ProfJenNeville !! Super important work.
Announcing the #ICLR2026 Outstanding Paper Awards 🏆 Congratulations to:
4
222
🇧🇷 I will be at ICLR 2026 in Rio! Come and say hi during our posters Thursday and Friday. For folks staying Sunday night that like futebol: Fluminense vs Chapecoense at Maracanã.
1
224
Bruno Ribeiro retweeted
🚀 Announcing RelBench V2, a major update to our benchmark for foundation models on relational data! With V2, we are significantly expanding the benchmark’s scope to catalyze further research in Relational Deep Learning (RDL) and Relational Foundation Models (RFMs). Key features: 🍺 4 new databases, spanning domains like e-commerce and beer reviews to scientific research and clinical healthcare. 🧩 40 new predictive tasks, including 28 autocomplete tasks, across new and existing databases. 🔌 External data integrations: 70+ datasets from CTU, 7 datasets from 4DBInfer, and your own data via SQL connector, all in RelBench format. 🛠️ Bug fixes and performance improvements. 🔥 Introducing autocomplete tasks: As opposed to forecasting tasks, autocomplete tasks predict existing columns in the database. We found that models need to deeply understand the relational context to autocomplete database fields, a critical capability that expands the scope of real-world RDL applications. Learn more: 🌐 Website: relbench.stanford.edu 💻 GitHub: github.com/snap-stanford/rel… Huge thanks to @justingu32 @_rishabhranjan_ @jakub_peleska @VHudovernik @CKanatsoulis @fengyuli607, Tang Haiming, Alistiq and everyone else who contributed to our GitHub for making this possible!
24
1
41
5,110
We propose a differentiable method to search over causal graphs that match the low-degree d-separation statements of the true causal graph. Come talk to us tomorrow in front of our poster #2407 from 11am-2pm.
4
14
2,105
🚀 Friday @ #NeurIPS2025 – We present the first unified causal‑discovery framework: percolation‑based DAG metrics + soft‑logic d‑separation bridge differentiable and constraint‑based methods. 🤝 Collaboration: @PurdueCS · @Ford · @JohnsHopkins #CausalDiscovery #Graphs #AI
Very excited to present our "Differentiable Constraint-Based Causal Discovery" at #NeurIPS2025! We introduce Differentiable d-Separation, bringing differentiable learning to Constraint-Based Causal Discovery. Come to chat and learn more! 📍 Hall C,D,E #2407 🕙 Fri 11am-2pm PST
6
854
Bruno Ribeiro retweeted
🤔 Is your data used to train machine learning models? 🔍 Membership Inference Attacks (MIA) are widely used to assess privacy risks of ML models. But existing MIAs all infer membership one instance at a time, ignoring their crucial dependence.
1
1
1
269