1/ One of most important things I've learned: Severe personality problems find *camouflage.* No one thinks “I’m a sadist” or “I'm a malignant narcissist.” They find a belief system/social group that validates their most hateful, destructive impulses & construes them as virtues.

🤝 Exciting news! HEREDITARY joins forces with the EU-funded project DataGEMS EOSC to revolutionize health data discovery, integration and exploration. 🔍 Read the full story on our website: hereditary-project.eu/strategic-syne… #HEREDITARYproject #DataGEMS #DataIntegration #HealthData

All I'm saying is we shouldn't fool ourselves. I'm very skeptical. E.g., what skills can you really learn from a single example other than, you know, surface the very skills that were developed intentionally at mid-training.

At this point, I basically treat anything Altman says about AI like a little kid bragging about how strong their imaginary friend is. Eh, scratch that. In both cases, there is nothing backing it up but the little kid thing is generally cuter.

Can you train a performant language models without using unlicensed text? We are thrilled to announce the Common Pile v0.1, an 8TB dataset of openly licensed and public domain text. We train 7B models for 1T and 2T tokens and match the performance similar models like LLaMA 1&2

Great (and unexpected) honor! I am also very thankful for this recognition!

📢 We're hiring a #PhD student at TU Wien for a project on Knowledge-Graph driven Factuality and Explainability in LLMs for conversational data analytics. 🌍 Funded by #MSCA #ARMADA_DN 📅 Apply by June 30 🔗 lnkd.in/dX2ZWtj8 #LLM #KnowledgeGraphs #Factuality #DataAnalytics

Poster Session 2 Wed, June 25 | 16:00–17:30 Lossless Transformation of Knowledge Graphs to Property Graphs using Standardized Schemas Kashif Rabbani, Matteo Lissandrini, Angela Bonifati, Katja Hose Kashif Rabbani M. Lissandrini Angela Bonifati Katja Hose

Deep learning training is a mathematical dumpster fire. But it turns out that if you *fix* the math, everything kinda just works…fp8 training, hyperparameter transfer, training stability, and more. [1/n]

Watch this space as we track who is a good (and bad) AI bot: goodaibots.com

This is true. This is sad. We should have done something about it, we should do something about it!

Nice to see a study on this! bukk.it/thisargumentis…

Vol:18 No:6 → PlanRGCN: Predicting SPARQL Query Performance vldb.org/pvldb/vol18/p1…

New position paper! Machine Learning Conferences Should Establish a “Refutations and Critiques” Track Joint w/ Sanmi Koyejo Joshua Kazdan Yegor Denisov-Blanch Francesco Orabona Koustuv Sinha Jessica Zosa Forde Jesse Dodge Susan Zhang Brando Miranda Matthias Gerstgrasser isha Elyas Obbad 1/6

It turns out the Turing Award is actually a silvery bowl from Tiffanys.

Did Winston Churchill coin the following popular saying? Courage is what it takes to stand up and speak; it is also what it takes to sit down and listen quoteinvestigator.com/2025/07/06/cou…

🔔VLDB 2025 Keynote Preview: Join Stratos Idreos from Harvard DASlab as he reimagines systems design in the AI era with his keynote on "Alphabets, Grammars, Calculators & the End of Hand-Crafted Systems". More details: vldb.org/2025/?program-… #VLDB2025 #DataSystems #AI

Adding two more open PhD positions at the department! Don't miss out on this opportunity to work with Jens Emil Grønbæk, Aslan Askarov, Stefanie Zollmann @[email protected], Tobias Langlotz, Ken Pfeuffer, Davide Mottin or Akhil Arora (@akhilarora.bsky.social) 🎉 Check out the projects and apply before Aug 1 ➡️ phd.nat.au.dk/for-applicants…

📣 Coming soon: a follow-up panel at VLDB 2025! With Shreya Shankar, Gustavo Alonso, Natacha Crooks, Jiannan Wang & Divesh Srivastava. We’ll dig into real trade-offs and concrete examples. Stay tuned.

Transactions on Graph Data and Knowledge, Volume 3, Issue 1 is now published! This issue includes two research articles, one survey article and one resource article, all available under Open Access for anyone to read at the following link: drops.dagstuhl.de/entities/issue…