Niv Cohen (@CohNiv) header
Niv Cohen (@CohNiv) profile picture

ECE faculty at the Technion

New York
Joined December 2020
Photo by Niv Cohen (@CohNiv)Photo by Niv Cohen (@CohNiv)Photo by Niv Cohen (@CohNiv)Photo by Niv Cohen (@CohNiv)Photo by Niv Cohen (@CohNiv)Photo by Niv Cohen (@CohNiv)Photo by Niv Cohen (@CohNiv)Photo by Niv Cohen (@CohNiv)Photo by Niv Cohen (@CohNiv)Photo by Niv Cohen (@CohNiv)
Piercing the veil to catch a glimpse of the Automated Grader.
25
128
10
1,865
35,597
Niv Cohen retweeted
1/ Today we are releasing Lea, a Lean theorem proving agent built for mathematicians. Lea has internally closed and formalized open problems in PDEs and theory CS, and comes out of DARPA's expMath program. 🧵
10
64
7
472
43,338
Niv Cohen retweeted
I'm proud to share the launch of Xcelsa Labs. The world is bottlenecked on compute. In order for the promise of AI to be delivered widely to the world, we need vastly better chips. (And we needed them yesterday). Siddharth Garg, Srikanth Jagabathula and I spent much of the last several months poring over how to make this happen. This is what we are doing at Xcelsa Labs (xcelsa.ai/); reach out if this is of interest!
5
1
25
2,856
Very happy to share one of the most exciting projects I've worked on, with this incredible group of people ✨ I'll be at USENIX Security and at CRYPTO next month (presenting this at the PPML workshop). DM me if you'll be there and want to talk about LLMs for cryptanalysis.
AI is advancing fast in math and cyber. Cryptanalysis sits at their intersection and underpins our digital security. So can LLMs do cryptanalysis? Increasingly, yes! Introducing CryptanalysisBench where we test if AI can break 191 real schemes from past standardization efforts
4
3
31
3,560
We tested several unlearning methods and found none of them really erase knowledge from the model - they simply hide it! 🧐 What does this mean? We must tread carefully with unlearning research within diffusion models🚨 Here is what we learned 🧵👇(led by @kevinlu4588)
Excited to share our paper “When Are Concepts Erased from Diffusion Models?” at @NeurIPSConf! We introduce two conceptual models for erasure mechanisms in diffusion models, and a suite of probes to recover supposedly forgotten concepts. Project website: unerasing.baulab.info/
1
14
46
2,405
Led by the amazing @kevinlu4588 !
Excited to share our paper “When Are Concepts Erased from Diffusion Models?” at @NeurIPSConf! We introduce two conceptual models for erasure mechanisms in diffusion models, and a suite of probes to recover supposedly forgotten concepts. Project website: unerasing.baulab.info/
5
358
If you are attending #ICML2025, check out our DataWorld workshop on Sat July 19. We have updated the website with more info on speakers & accepted papers! dataworldicml2025.github.io/ Also happy to chat offline about all things ✨ data ✨
25
2
81
11,541
🎤 Giving a keynote talk - "Anomaly Detection Beyond Pretrained Representations" tomorrow at 10:30am at #CVPR, VAND 3.0 workshop (Visual Anomaly and Novelty Detection). Come say hi! 👋
1
2
1
35
6,064
When we "erase" a concept from a diffusion model, is that knowledge truly gone? 🤔 We investigated, and the answer is often 'no'! Using simple probing techniques, the knowledge traces of the erased concept can be easily resurfaced 🔍 Here is what we learned 🧵👇
2
8
1
33
1,820
📣We are extending our deadline to May 31st!📣 Looking forward to seeing everyone's submissions :)
📢 Announcing our data-centric workshop at ICML 2025 on unifying data curation frameworks across domains! 📅 Deadline: May 24, AoE 🔗 Website: dataworldicml2025.github.io/ We have an amazing lineup of speakers + panelists from various institutions and application areas.
4
7
864
📢 Announcing our data-centric workshop at ICML 2025 on unifying data curation frameworks across domains! 📅 Deadline: May 24, AoE 🔗 Website: dataworldicml2025.github.io/ We have an amazing lineup of speakers + panelists from various institutions and application areas.
2
26
5
133
26,015
Think your latent-noise diffusion watermarking method is robust? Think again! We show that they are susceptible to adversarial attacks that only require one watermarked example and an off-the-shelf encoder. This attack can forge and remove the watermark with very high accuracy
1
10
1
30
2,769
I'm excited about the recent feature we've been working on at @pika_labs - action editing! 🏃 📹 Video editing methods often focus on spatial edits. Our method can directly edit the *action* of objects, even when the video already contains a distinctive motion (e.g, 👋).
19
74
10
715
82,913
In our #ICLR2025 paper, we introduce WIND 🌬️ A method that embeds a distortion watermark directly in the diffusion noise! Our method ensures that the watermark in one image does not reveal information about the watermark in other images 🤫 📝 arxiv.org/abs/2412.04653 (1/5)
1
20
1
64
5,584
And there's more! By leveraging diffusion inpainting, WIND can also protect non-AI-generated images, opening exciting new pathways for copyright protection and authenticity verification across all digital content 🖼️🔑 (4/5)
1
194
In summary, WIND delivers robust, imperceptible watermarking for both AI-generated and natural images 🛡️🔏 For detailed results and insights, see our paper or code: github.com/Kasraarabi/Hidden… Led by Kasra Arabi in collaboration with @FeuerBenjamin @TealWitter @chegday (5/5)
2
220
Niv Cohen retweeted
Today we launch a new open research community with @wattenberg @viegasf @jacobandreas Ellie Pavlick @Brown_NLP and many others It is called ARBOR. please join us. 🧵
1
14
2
36
5,215