Nir Goren nag-retweet
We developed a simple, sample-efficient online RL technique for post-training image generation models. We see it as a possible steerable alternative to CFG, driven by any scalar reward, including human preference.
English
Nir Goren
245 posts
Nir Goren
@nirgoren
Researching AI https://t.co/eMCNT73s4T
Sumali Mayıs 2018
117 Sinusundan115 Mga Tagasunod
English
English
CVPR 2026 highlight! 🔥
In this work co-led with @YehezkelShai, we show that a plain diffusion model can solve hard geometry problems by treating them as conditional image generation problems. No special architecture needed.
w/ @OmerDahary, @kusichan, @OPatashnik, @DanielCohenOr1
Shai Yehezkel@YehezkelShai
Visual Diffusion Models are Geometric Solvers We cast geometry as images: a plain diffusion model denoises into valid solutions. It is simple, general and effective. Shown on Inscribed Square, Steiner Tree, and Maximum Area Polygonization - all classic hard problems.
English
Nir Goren nag-retweet
[1/5] Is Text Enough for Control? 🐇
Text-driven video editing lets you describe *what* to change. But what about *how much*?
We introduce Adaptive-Origin Guidance (AdaOr).
A joint work with @DecartAI and @TelAvivUni 🧪
accepted to #SIGGRAPH2026.
English
Nir Goren nag-retweet
Video models as Physics simulators. 🌍🎥
[1/] In our latest work, WinDiNet, we finetuned a pre-trained video model into a differentiable physics engine. 1000x faster than traditional CFD solvers.
Project page: rbischof.github.io/windinet_web/
Abs: arxiv.org/abs/2603.21210
English
Nir Goren nag-retweet
Our previous intern released an extremely impressive re-implemented demo of our paper on multiplayer diffusion game engines.
play-multigen.com
I think this might be the first time you can play a fully-functional multiplayer generative game online with other people. 🤯
English
Nir Goren nag-retweet
Modern T2I DiTs are incredibly powerful, but have a serious diversity problem.
We introduce a surprisingly simple and efficient inference-time fix
(+2s for Flux-dev, +1s for SD3.5-Turbo).
Excited to share our SIGGRAPH 2026 (conditional) paper:
“On-the-fly Repulsion in the Contextual Space for Rich Diversity in Diffusion Transformers”
English
Nir Goren nag-retweet
Excited to share that our work, “Image Generation from Contextually Contradictory Prompts,” was accepted to CVPR 2026 🎉
Diffusion models fail on some seemingly simple prompts.
Why do they ignore what you asked for?
We show why, and how to fix it with a simple, training-free method.
Joint work with @OPatashnik @OmerDahary @MokadyRon @DanielCohenOr1
English
Nir Goren nag-retweet
DLSS 5 is all over the timeline, and for good reason. In my internship at @AIatMeta we had the same idea: use a video model as a learned second-stage renderer on top of game engines. In our paper RealMaster, we make synthetic video look real while preserving scene fidelity 👇
English
Nir Goren nag-retweet
Video editing just got more dynamic! 🚀
Thrilled to share DynaEdit: a training-free, text-based method for non-rigid video editing.
Work done during my internship at @GoogleDeepMind with @Roni_Paiss, @kusichan, @inbar_mosseri, @talidekel, @t_michaeli
dynaedit.github.io
English
@PolaczekSagi {
"hooks": {
"Notification": [
{
"matcher": "",
"hooks": [
{
"type": "command",
"command": "printf '\\a' > $(tmux display-message -p '#{pane_tty}')"
}
]
}
]
}
English
@PolaczekSagi btw, you can set up ~/.claude/settings.json for a terminal bell on completion/permission request that works over claude code in tmux (which you can turn into a system notification on iTerm2 when not in focus):
English
iTerm2 + 'tmux -CC'
best settings for persistent remote development I've been using so far. easily recovers remote sessions (e.g. claude) even when get disconnected - which happens often in the bomb shelters 👀
does anyone have a better one?
English
Nir Goren nag-retweet
Nir Goren nag-retweet
SemanticMoments - Semantic motion similarity
How do you find videos with similar motion?
It’s harder than it sounds.
Models like VideoMAE and V-JEPA encode motion, but their embeddings are dominated by appearance.
So how do we build a compact embedding for motion similarity?
Joint work with @kfir99 @OPatashnik @BenaimSagie @MokadyRon
GIF
English
English
@nirgoren @KatzirOren @abhinav_nakarmi @eyalr0 @mahmoods01 @OPatashnik Such an interesting paper! Congrats Nir! 👏
English
Happy to share that our paper "NoisePrints: Distortion-Free Watermarks for Authorship in Private Diffusion Models" has been accepted to ICLR 2026!
Huge thanks to my amazing co-authors:
@KatzirOren @abhinav_nakarmi @eyalr0 @mahmoods01 @OPatashnik
#ICLR2026 🇧🇷
Nir Goren@nirgoren
The initial noise in diffusion models is surprisingly correlated with the final image. Our NoisePrints paper exploits this to provide a lightweight, distortion-free, cryptographically secure watermark for proving authorship of generated images & videos, requiring no model access.
English
Nir Goren nag-retweet
🚀 LTX-2 is now open source: text → audio + video.
Today we’re releasing LTX-2, the first open-source foundation model for joint audiovisual generation, together with a full technical report.
🧵👇
English
Nir Goren nag-retweet
[1/4] Sync about it… 💭✨
Editing a portrait video yet keeping it fully synced with the original across the entire sequence.
Read more about Sync-LoRA: sagipolaczek.github.io/Sync-LoRA/ 🚀
English
Might be interesting to see how see training a diffusion network on the visual representation of ARC fares. We demonstrated that visual diffusion models are capable reasoners on some other hard tasks when they are represented as images: arxiv.org/abs/2510.21697
Rosinality@rosinality
You can just train ViT from scratch to solve ARC.
English