Ravid Shwartz Ziv @ICML
@ziv_ravid
Faculty Fellow and Assistant Professor at @NYUDataScience. Talk to me about LLMs, Compression and Tabular Data
@ylecun and I have been pondering the concept of optimal representation in self-supervised learning, and we're excited to share our findings in a recently published paper! 📝🔍 arxiv.org/abs/2304.09355
Will AI Take Our Spiritual 'Job'? AI, Torah, and the Future of Human Understanding. Can AI unlock the structured knowledge of the Torah corpus? Can the Torah's comprehensive framework transform our approach to AI? Featuring @harryritter1, @CLeibowicz, @ziv_ravid,…
הי טוויטר, אני מגיע לבקר בישראל שבוע הבא לחודש 🥳🥳🥳 אם אתם רוצים לדבר על מודלי שפה, אג'נטים, דחיסה וזיכרון או בכלל אקדמיה, תעשיה וסטארטפים אני אשמח להיפגש.
Qwen3 finally separated their hybrid thinking mode into separate instruction and thinking models, achieving SOTA results. No more hybrid chat template🥳🥳🥳
Bye Qwen3-235B-A22B, hello Qwen3-235B-A22B-2507! After talking with the community and thinking it through, we decided to stop using hybrid thinking mode. Instead, we’ll train Instruct and Thinking models separately so we can get the best quality possible. Today, we’re releasing…
I just came back from ICML. Thank you, everyone; it was great to meet with you all and hear so many new and exciting ideas!
A moment of bragging! If you look at Whova to see which orals/posters had the highest likes, it turns out to be our layer-by-layer paper 🥳 Did we develop a multiagent reasoning system that hacked the app and added fake attendees? You'll never know 🤫
Someone on LinkedIn posted about cool theoretical research that he wants to check, and someone from AMD just told him that they will give him the compute 😍
Well played Noam, well played...
It takes us a few months to turn the experimental research frontier into a product. But progress is so fast that a few months can mean a big difference in capabilities.
All the ICML company booths are giving out hats. At least when the AI bubble bursts, we'll all be ready to play golf 🥳🥳🥳
Oh, you're an AI startup founder? Let me guess... Don't tell me - you're looking for AI engineers who actually know how to build things? And let me guess again - they're incredibly rare and you just can't seem to find them? No way!
I totally agree. There are good ideas all the time, and many talented people who can implement them. The culture and the decision process are the fundamental differences between the teams
It’s funny that people on this site think major LLM efforts are talent-bound rather than org-bound. The talent differential has never been big between major orgs. Most of the difference in outcomes is due to organisational factors - like allocating compute to the right bets, and…
The best thing about conferences is talking with people and hearing new ideas. I just talked with someone with such an elegant and great idea about agents that I was jealous that I didn't think about it myself 😄
On my way to #ICML2025. Reach out if you want to talk!
Do you want to speed up your inference time by 2X for free? Using speculative decoding (SP) but want to use your own drafter? Want lossless speedup with one line? Go to @NadavTimor talk at ICML tomorrow at 3:30 PM and check out his poster afterward at 4:30 PM. Thank me later!
If you're at ICML, come tomorrow (Tuesday) to Oscar's talk, where he will present our paper "Layer by layer: Uncovering hidden representations in language models" at 10am (West Ballroom D) and for the poster session at 11am (East Exhibition Hall A-B #E-2607).