Hady Elsahar
LightNews LightNews
hadyelsahar.bsky.social
Hady Elsahar
@hadyelsahar.bsky.social
14 followers 18 following 10 posts
Senior Research Scientist at FAIR (Meta)

hadyelsahar.io
Posts Replies Media Videos
Reposted by Hady Elsahar
hardmaru.bsky.social
New Paper: Continuous Thought Machines

pub.sakana.ai/ctm/

Neurons in brains use timing and synchronization in the way that they compute, but this is largely ignored in modern neural nets. We believe neural timing is key for the flexibility and adaptability of biological intelligence.

Thread ↓
May 12, 2025 at 2:38 AM
Reposted by Hady Elsahar
iclr-conf.bsky.social
That's a wrap for #ICLR2025! See you all next year in Brazil! Please all welcome @bharathhariharan.bsky.social as the new Senior Program Chair! (With @cvondrick.bsky.social continuing on as General Chair.)
April 28, 2025 at 10:34 AM
hadyelsahar.bsky.social
Furong Huang @furongh.bsky.social now giving a talk about the waves benchmark - robustness of image watermarking at the wmark workshop #iclr2025
April 27, 2025 at 2:02 AM
hadyelsahar.bsky.social
Scott Aaronson is starting with a keynote talking about his work in the super alignment team at OpenAI #wmark #iclr2025
April 27, 2025 at 1:16 AM
hadyelsahar.bsky.social
60 submissions to the watermarking workshops at #iclr2025 half of them is about text watermarking
April 27, 2025 at 1:10 AM
Reposted by Hady Elsahar
iclr-conf.bsky.social
Loving this poster session energy! #ICLR2025
April 24, 2025 at 5:34 AM
hadyelsahar.bsky.social
Still don’t know which workshop to attend at #ICLR2025?

Come by WMARK — the 1st Workshop on GenAI Watermarking!
With 51 papers, amazing speakers,watermarking is becoming one of the most important and timely topics in AI.

Maybe your next research idea starts here!
April 26, 2025 at 11:50 PM
Reposted by Hady Elsahar
abeirami.bsky.social
Today at the 10am #ICLR2025 poster session, I will present Ziteng Sun's paper "block verification accelerates speculative decoding"
April 24, 2025 at 9:50 PM
hadyelsahar.bsky.social
Just met Chris Manning he said “we seem to be doing well” The younger me would have been proud hearing this. #iclr2025
April 24, 2025 at 10:01 AM
hadyelsahar.bsky.social
@zicokolter.bsky.social thinks AI safety is a research topic to relive the iclr 2015 moment again and secure a new lasting impact , I tend to agree :) #ICLR2025
April 24, 2025 at 1:52 AM
hadyelsahar.bsky.social
@zicokolter.bsky.social on what went wrong that prevented the adoption of certified robustnesss #iclr2025
April 24, 2025 at 1:30 AM
Reposted by Hady Elsahar
yoshitomo-matsubara.net
[Please share🔄]

Did you read blog posts / preprint about #ICLR2025 LLM experiment?
Were you in the #ICLR2025 review process?

As I was not in the process, I'm asking for opinions to improve review systems

Thank you for your help🙏

🤗 huggingface.co/blog/yoshito...
Personal thoughts on a randomized study of LM-based review feedback agent at ICLR 2025
A Blog post by Yoshitomo Matsubara on Hugging Face
huggingface.co
April 22, 2025 at 3:59 PM
hadyelsahar.bsky.social
Heya! First conference experience on Bsky 🦋

I'm at #ICLR2025 in Singapore 🇸🇬 this week organizing the 1st Workshop on GenAI Watermarking (WMARK) on April 27 — packed day of talks, posters, and a great panel 👇

And check out our new paper on watermarking taxonomy 👉 openreview.net/pdf?id=4kKLL...
A flyer for the “1st Workshop on GenAI Watermarking (WMARK@ICLR)” held on April 27th, 2025, from 9am to 5pm at ICLR 2025, Singapore Expo, Hall 4. The flyer includes a title, schedule, speaker photos, and a description. The tagline reads, “A Dedicated Space for Watermarking in Generative AI,” explaining the importance of watermarking for AI safety and security. Speakers & Panelists: Photos and names of the speakers are shown in a grid format: Scott Aaronson Furong Huang John Collomosse Mauro Barni Melissa Omimo Adina Yakefu Lee Wan Sie Zohaib Ahmed Schedule: Morning Session 09:00–09:10: Introduction 09:10–09:40: Scott Aaronson – “Watermarking, Where to?” (Invited Talk) 09:45–10:15: Furong Huang – “Erasing the Invisible” (Invited Talk) 10:15–10:30: Coffee Break 10:30–10:45: Talk on “Cryptography in Semantic Watermarks for Diffusion Models” 10:45–11:00: Talk on “Adaptive Attacks against Content Watermarks for Language Models” 11:00–12:00: Poster Session A 12:00–13:30: Lunch Break Afternoon Session 13:30–14:30: Poster Session B 14:30–15:00: Melissa Omimo – “Watermarking in GenAI: The Case of African Voices” (Invited Talk) 15:00–15:15: Coffee Break 15:15–15:45: John Collomosse – “Building Safe and Fair Generative AI with Content Provenance” (Invited Talk) 15:45–16:15: Mauro Barni – “Dos and Don’ts: Lessons from 30 Years in the Field” (Invited Talk) 16:15–17:15: Panel Discussion featuring Zohaib Ahmed, Adina Yakefu, Lee Wan Sie, and John Collomosse 17:15–17:30: Closing Remarks A QR code is included at the bottom left linking to https://sites.google.com/view/genai-watermark. Background includes a city skyline and water reflections.
April 23, 2025 at 10:54 PM
hadyelsahar.bsky.social
🚀 Call for Papers! Join us at WMark at #ICLR2025 (@iclr-conf.bsky.social) —a dedicated workshop on Watermarking in GenAI.

🔥 Amazing speakers + policy experts!
📆 Submit by Feb 3, 2025
✅ Tiny papers welcome
✅ Dual submissions allowed.

Details: sites.google.com/view/genai-w...
An image showcasing the "Speakers & Panelists" for the 1st Workshop on GenAI Watermarking (WMark @ ICLR2025). The layout features photos and brief descriptions of seven distinguished speakers and panelists, including: Mauro Barni, professor at the University of Siena with expertise in digital watermarking. John Collomosse, professor at the University of Surrey and Adobe scientist focused on content provenance. Scott Aaronson, chair of Computer Science at the University of Texas, currently on leave at OpenAI. Tom Goldstein, professor at Maryland University known for influential work in AI security and privacy. Zohaib Ahmed, CEO of Resemble AI, focusing on AI voice cloning with an emphasis on consent and transparency. Melissa Omino, director at the Center for Intellectual Property and Information Technology Law (CIPIT). Adina Yakefu, executive assistant at Hugging Face, contributing to the growth of the ML community in China. The image is arranged in a clean, grid-like format with each speaker’s name linked and accompanied by their credentials.
January 8, 2025 at 10:10 AM