Agrim Gupta
@agrimgupta92
Followers
4K
Following
619
Media
37
Statuses
343
Simulating reality @GoogleDeepMind, prev PhD @Stanford
Joined January 2017
We introduce W.A.L.T, a diffusion model for photorealistic video generation. Our model is a transformer trained on image and video generation in a shared latent space. 🧵👇
51
249
1K
RT @SeanKirmani: 🤖🌎 We are organizing a workshop on Robotics World Modeling at @corl_conf 2025!. We have an excellent group of speakers and….
0
35
0
RT @keshigeyan: 1/ Model architectures have been mostly treated as fixed post-training. 🌱 Introducing Grafting: A new way to edit pretrain….
0
43
0
RT @RuiqiGao: 📢 We are organizing a workshop at ICML on Building Physically Plausible World Models. Join us by submitting your cool papers….
0
13
0
RT @SeanKirmani: 🌎🌏🌍 We are organizing a workshop on Building Physically Plausible World Models at @icmlconf 2025!. We have a great lineup….
0
22
0
RT @abhisk_kadian: Meet Llama4 Scout & Maverick: the next generation of Llama 🦙 🚀. Download: Blog: .
0
3
0
RT @SuryaGanguli: My @TEDAI2024 talk is out!.I discuss our work, spanning AI, physics, math & neuroscience, to deve….
0
46
0
RT @avdnoord: Our image model is on LMSYS : ). It's been an amazing effort by the team, I'm very proud of what we achieved over the last ye….
0
46
0
RT @SuryaGanguli: Nice name! We can share :).MetaMorph: Learning Universal Controllers with Transformers.
0
2
0
RT @RubenEVillegas: A cat roars while looking at its reflection in the mirror but instead sees itself as a lion roaring #veo2
https://t.co/….
0
209
0
🤯.
1
3
55
RT @GoogleDeepMind: Today, we’re announcing Veo 2: our state-of-the-art video generation model which produces realistic, high-quality clips….
0
1K
0
Today we are introducing Veo2: SOTA video generation model. You can try the model here:
Introducing Veo 2, our new, state-of-the-art video model (with better understanding of real-world physics & movement, up to 4K resolution). You can join the waitlist on VideoFX. Our new and improved Imagen 3 model also achieves SOTA results, and is coming today to 100+ countries
3
3
129
RT @jparkerholder: Introducing 🧞Genie 2 🧞 - our most capable large-scale foundation world model, which can generate a diverse array of cons….
0
475
0
RT @drfeifei: Super proud of this work by my student @keshigeyan and our collaborators on pushing the benchmarking of LLMs for spatial and….
0
50
0
RT @ManlingLi_: HourVideo: New benchmark for hour-long egocentric videos!. Most challenging benchmark so far (SOTA 37.3% → Humans 85.0%). A….
0
7
0
New benchmark for long video understanding. LLMs have made significant strides in reasoning, handling long contexts, and solving complex quantitative tasks. But they still struggle with something humans find “easy”: processing visual information over extended time periods. Our.
1/ [NeurIPS D&B] Introducing HourVideo: A benchmark for hour-long video-language understanding!🚀. 500 egocentric videos, 18 total tasks & ~13k questions!. Performance:.GPT-4➡️25.7%.Gemini 1.5 Pro➡️37.3%.Humans➡️85.0%. We highlight a significant gap in multimodal capabilities🧵👇
1
7
31
RT @DrJimFan: Not every foundation model needs to be gigantic. We trained a 1.5M-parameter neural network to control the body of a humanoid….
0
508
0
RT @abhisk_kadian: Llama3.2 models are here 🎉! We are releasing the multimodal and lightweight Llama models.
0
5
0
RT @jcjohnss: Spatial Intelligence lets us perceive, reason, act, and create in the rich 3D world around us. Today we are launching World L….
0
26
0