🔥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

For those who missed it in the crowds yesterday: a shot of our poster on "The Price of Implicit Bias" with Nikolaos Tsilivis, Natalie Frank and Nati Srebro. NYU Center for Data Science AI at Meta with a nice writeup here: nyudatascience.medium.com/robustness-at-…

For those into jailbreaking LLMs: our poster "Mission Impossible" today shows the fundamental limits of LLM alignment - and improved ways to go about it, nonetheless. With Dr. Karen Ullrich & Jingtong Su #2302 11am - 2pm Poster Session 3 East NYU Center for Data Science AI at Meta #NeurIPS2024

Interested in implicit bias of algorithms like Adam and Shampoo? Pass by our poster "Flavors of Margin" today (Sat) at the M3L Workshop @ NeurIPS 2024 4-5pm, East Meeting Room 1-3 with Nikolaos Tsilivis & gal vardi NYU Center for Data Science AI at Meta Weizmann Institute based on arxiv.org/abs/2410.22069

If you - like many - believe that using more (good) data is better than repeating on less - guess again! And come to our poster "Emergent Properties..." at the SciForDL Scientific Methods for Understanding Deep Learning workshop West Mtg Room 205-207 this afternoon (after 4pm)! With François Charton AI at Meta NYU Center for Data Science

Congratulation to the winners of our debunking challenge François Charton and Julia Kempe for "Emergent properties with repeated examples" 🥳🎉

Thanks to co-author & colleague François Charton for this collaboration (Emergent Properties w Repeated Examples) that won the Debunking Challenge at SciforDL Scientific Methods for Understanding Deep Learning #NeurIPS2024 . We both learned that - at least sometimes - one epoch is *not* all you need AI at Meta NYU Center for Data Science

Submit to our New Frontiers in Associative Memories workshop ICLR 2026. New architectures & algorithms, memory-augmented LLMs, energy-based models, Hopfield networks, assoc. memory & diffusion.. nfam.vizhub.ai openreview.net/group?id=ICLR.… Organizing with Dmitry Krotov et al

Model collapse & AI data: Happy to work with BBC Radio4 in this podcast bbc.co.uk/sounds/play/m0… Impressed by thoughtful questions and moderation of Aleks Krotoski & Dr Kevin Fong Features our work with Elvis Dohmatob Yunzhen Feng Pu Yang 杨 埔 François Charton at NYU Center for Data Science and AI at Meta

AI bias grows when data pruning favors simplicity over fairness. CDS PhD graduate Artem Vysogorets (Artem Vysogorets) and CDS Silver Professor Julia Kempe (Julia Kempe) propose DRoP—a method boosting fairness and worst-case accuracy by up to 10%. Learn more: nyudatascience.medium.com/drop-making-ai…

PILAF (Policy-Interpolated Learning for Aligned Feedback): our response sampling scheme that provably aligns LLM preference learning w maximizing the underlying oracle reward! arxiv.org/abs/2502.04270 Yunzhen Feng Ariel Kwiatkowski Kunhao Zheng Yaqi Duan AI at Meta NYU Center for Data Science

You think on-policy sampling gives the best reward models? Think again! 🔥 Our finding: Even with on-policy data, reward models misalign with policy optimization goals! Introducing PILAF—strategic sampling that fixes this fundamentally. (1/11)

It is a real delight to work with Elvis Dohmatob and I encourage every student in search of excellent and rigorous mentorship to apply to his group!

We refused to cite the paper due to severe misconduct of the authors of that paper: plagiarism of our own prior work, predominantly AI-generated content (ya, the authors plugged our paper into an LLM and generated another paper), IRB violations, etc. Revealed during a long

Julia Kempe 😆

If in Singapore next week, come by our #ICLR2025 Spotlight poster for our recent study at Julia Kempe unveiling how data pruning promotes implicit bias in datasets and proposing a method (DRoP) that does exactly the opposite: iclr.cc/virtual/2025/p…

Here is to a next generation of AI-literate kids! International AI Olympiad ioai-official.org ML Researchers, you might appreciate the impressive syllabus. Do we have all the chops our kids are expected to have :) ? ioai-official.org/wp-content/upl…

Check out our poster tmr at 10am at the ICLR Bidirectional Human-AI Alignment workshop! We cover how on-policy preference sampling can be biased and our optimal response sampling for human labeling. NYU Center for Data Science AI at Meta Julia Kempe Yaqi Duan x.com/feeelix_feng/s…

Congrats to 37 CDS researchers — faculty, postdocs, and PhD students — who had papers accepted to ICLR 2025, including Spotlighted work by Julia Kempe, Yunzhen Feng, Andrew Gordon Wilson, Yilun Kuang, and Jianyu Zhang. Full list: nyudatascience.medium.com/cds-researcher…