moab.arar Profile
moab.arar

@ArarMoab

Followers
243
Following
363
Statuses
183

Ph.D. student | Generative Models

Joined March 2021
Don't wanna be here? Send us removal request.
@ArarMoab
moab.arar
5 months
Checkout our work "GameNGen". A Gaming engine powered by a diffusion-model that simulates DOOM in Real-Time! Find out more: Amazing effort and fun collaboration with the incredible @daniva, @yanivle, and @shlomifruchter!
@_akhaliq
AK
6 months
Google presents Diffusion Models Are Real-Time Game Engines discuss: We present GameNGen, the first game engine powered entirely by a neural model that enables real-time interaction with a complex environment over long trajectories at high quality. GameNGen can interactively simulate the classic game DOOM at over 20 frames per second on a single TPU. Next frame prediction achieves a PSNR of 29.4, comparable to lossy JPEG compression. Human raters are only slightly better than random chance at distinguishing short clips of the game from clips of the simulation. GameNGen is trained in two phases: (1) an RL-agent learns to play the game and the training sessions are recorded, and (2) a diffusion model is trained to produce the next frame, conditioned on the sequence of past frames and actions. Conditioning augmentations enable stable auto-regressive generation over long trajectories.
1
17
49
@ArarMoab
moab.arar
6 days
RT @abk_tau: DeciMamba, the first context extension method for Mamba, is accepted to #ICLR2025! 🎉 New revision with more long-context resu…
0
6
0
@ArarMoab
moab.arar
6 days
RT @hila_chefer: VideoJAM is our new framework for improved motion generation from @AIatMeta We show that video generators struggle with m…
0
191
0
@ArarMoab
moab.arar
8 days
I just came across this nice paper - it turns out (some) multi-modal models look at images through the lens of the text tokens!
@omri_kaduri
Omri Kaduri
2 months
🔍 Unveiling new insights into Vision-Language Models (VLMs)! In collaboration with @OneViTaDay & @talidekel, we analyzed LLaVA-1.5-7B & InternVL2-76B to uncover how these models process visual data. 🧵
0
0
3
@ArarMoab
moab.arar
10 days
Choose you co-authors wisely ✍️
@ICCVConference
#ICCV2025
11 days
Check out the changes for #ICCV2025 🌶️
Tweet media one
0
0
3
@ArarMoab
moab.arar
12 days
Interesting application and cool results
@AbdalRameen
Rameen Abdal
12 days
What if you could compose videos— merging multiple clips, even capturing complex athletic moves where video models struggle - all while preserving motion and context? And yes, you can still edit them with text after! Stay tuned for more results. #AI #VideoGeneration #SnapResearch
0
0
2
@ArarMoab
moab.arar
12 days
@OmriAvr @pika_labs Looking sharp! Just like you did at CVPR ‘22 😃 Good luck!
0
0
5
@ArarMoab
moab.arar
14 days
@EladRichardson Awesome as always!
0
0
1
@ArarMoab
moab.arar
14 days
@CSProfKGD @DJIGlobal Your students are lucky - you put alot into teaching!
1
0
4
@ArarMoab
moab.arar
14 days
@OmriAvr Wow 🤩 this looks amazing
0
0
0
@ArarMoab
moab.arar
18 days
RT @lipmanya: Our **Flow Matching Tutorial** from #NeurIPS2024 is now publicly available: @helibenhamu @RickyTQCh
0
92
0
@ArarMoab
moab.arar
19 days
GameNGen has been accepted to #ICLR2025! 🎉 Huge congrats to my incredible co-authors @daniva, @yanivle, and @shlomifruchter—it was an amazing effort and such a fun collaboration! Learn more:
@_akhaliq
AK
6 months
Google presents Diffusion Models Are Real-Time Game Engines discuss: We present GameNGen, the first game engine powered entirely by a neural model that enables real-time interaction with a complex environment over long trajectories at high quality. GameNGen can interactively simulate the classic game DOOM at over 20 frames per second on a single TPU. Next frame prediction achieves a PSNR of 29.4, comparable to lossy JPEG compression. Human raters are only slightly better than random chance at distinguishing short clips of the game from clips of the simulation. GameNGen is trained in two phases: (1) an RL-agent learns to play the game and the training sessions are recorded, and (2) a diffusion model is trained to produce the next frame, conditioned on the sequence of past frames and actions. Conditioning augmentations enable stable auto-regressive generation over long trajectories.
6
5
36
@ArarMoab
moab.arar
20 days
@m__dehghani @alexanderchen @AniBaddepudi @riedelcastro Awesome work and presentation!
0
0
1
@ArarMoab
moab.arar
22 days
@DrJimFan Will models ever achieve true (genius-level) creativity? Only then could a transformer invent the next generation of itself.
0
0
2
@ArarMoab
moab.arar
25 days
Some Med Schools understand it, and you can see adaptation in their Curriculum. Also I think human validation would still be needed - so the medical specialists will not be replaced but will need to adapt. Full autonomous AI doctor is not around the corner.
@DeryaTR_
Derya Unutmaz, MD
26 days
The medical specialties most likely to be replaced by AI soon, in order: Pathology Radiology Ophthalmology (non-surgical) Dermatology (non-intervention) Primary Care Internal medicine (diagnostics) In 10 years, 90% of medical specialties will be replaced by AI+robots
0
0
1
@ArarMoab
moab.arar
27 days
Finally, the field has grown at a crazy pace, and you may find yourself discussing dozens of papers. Try to group approaches into common categories, highlight their strengths and weaknesses and explain why your method is interesting.
0
0
0