🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

Great to have one of our own USC ECE grads* Dimitris Papailiopoulos back for an awesome talk on teaching algorithms via self improvement. More detail sites.usc.edu/aif4s/events/

I hear is getting a tad bit chilly 🥶 this weekend for some folks. Friendly reminder that u can make better life choices usccareers.usc.edu/job/los-angele… Yes this is January, yes this is an outdoor pool, no it is not heated, and yes it will be part of the 2028 LA Olympics!

Great to see others discovering similar findings as we did in our Neurips2024 paper (arxiv.org/abs/2406.03445). We call these Fourier features instead of helix. How are these features useful for representing numbers? Stay tuned for our new number embedding paper coming soon!

Our work (with Tianyi Zhou Deqing Fu and Vatsal Sharan) published at NeurIPS 2024 already found that pretrained LMs do addition using modular arithmetic/“trigonometry” (we called these Fourier features). Indeed it is a clever mechanism.

Billion-parameter LLMs still struggle with simple arithmetic? 📞 FoNE (Fourier Number Embedding) tackles this problem. By mapping numbers directly into Fourier space, it bypasses tokenization and significantly improves numerical accuracy with better efficiency and accuracy.

Exciting times for AI in bio two impressive model releases yesterday Evo 2 and bioemu 1

Opt friends: If I have PL inequality +lipshitzness instead of smoothness do we have convergence guarantees for GD updates. Obviously don’t expect geometric rates. In general good refs for convergence guarantees for PL without smoothness would be appreciated. Cc Damek

Quiet moments above LA after another rewarding day at #AAR25. A time to reflect, reset, and look ahead. 📸 Mahdi Soltanolkotabi

Sub 21 rounds down? strava.app.link/MvFj7G6qXRb

Our 2025 RSS workshop on "Statistical Uncertainty Quantification in the Era of AI-Enabled Robots" got accepted. We have an amazing lineup of tentative speakers, see sites.google.com/view/rss2025-w… 🚀 The workshop will be held at USC on June 25th (and we guarantee excellent weather 🏖️🌴)

Looking forward to this workshop at Halıcıoğlu Data Science Institute, the lineup looks amazing. Registration is open.

Ok I’ve seen enough I hereby give Phi4 Esfahan’s stamp of approval 😅

Friendly reminder that small random noise has many benefits: improves convergence, reliability, generalization, privacy & reduces deadline conflicts. This Thu I have 3 separate deadlines at 9AM, 12 PM, the AOE one, and end of semester craziness. All docs on overleaf 🤦‍♂️

Plan on doing some random summer reading. Pls send your favorite “Agents” papers by you or others (if possible with a short explanation of why u think is a good paper)

If you have written/seen a cool new paper on post training / RL / agents, please share a link and tl;dr. I'm looking for some new reading material.

Happy to see the nature in Pacific Palisades slowly healing from the fires. I recently joined the school of eng working group on the LA fires. If anyone has any good ideas re where research efforts could be most useful for rebuilding/future prevention pls reach out

We're in an Evaluation crisis!

Big news! USC's new Institute on AI and Ethics will be co-directed by USC Thomas Lord Department of Computer Science Professor Yan Liu with USC Dornsife College of Letters, Arts & Sciences's Professor John Hawthorne 👏 USC Viterbi School USC School of Advanced Computing

After three incredible years, today is my last day at Google DeepMind! I am truly grateful to the amazing colleagues who made the journey 1000x more fruitful and enjoyable! I am forever indebted to my collaborators who showed me how to be better at everything via demonstrations.

Great, detailed explainer. Indirect costs ARE research costs, but are ones that are hard to assign to individual projects. x.com/pottytheron/st…