Video yükleniyor...

Video Yüklenemedi

Bu video yüklenirken bir sorun oluştu. Bu geçici bir ağ sorunundan kaynaklanıyor olabilir veya video kullanılamıyor olabilir.

Ana Sayfaya Dön

Diffusion/flow matching is a go-to generative AI paradigm, but it's slow, sample inefficient and complicated. Do we really need all this complexity? We design a small and simple one-step model, ROMS-IMLE, which attains an FID of 2.56 on ImageNet. Joint w/ Chirag Vashist 1/6

Ke Li 🍁

7,395 subscribers

37,380 görüntüleme • 1 ay önce •via X (Twitter)

Bilim & Teknoloji

Anya Rossi• Live Now

Private livecam show

0 Yorum

Yorum bulunmuyor

Orijinal gönderinin yorumları burada görünecek

Benzer Videolar

Diffusion and flow matching-based robot planners are slow and generate noisy and jerky trajectories. Delighted to share our ICRA 2026 paper, which leverages IMLE to improve planning frequency 19-fold from 4.3 Hz to 83 Hz and reduces jerk by 38% relative to flow matching. Joint work w/ Grayson Lee, Minh Bui, Shuzi Zhou, Yankai Li and Mo Chen. (1/7)

Diffusion and flow matching-based robot planners are slow and generate noisy and jerky trajectories. Delighted to share our ICRA 2026 paper, which leverages IMLE to improve planning frequency 19-fold from 4.3 Hz to 83 Hz and reduces jerk by 38% relative to flow matching. Joint work w/ Grayson Lee, Minh Bui, Shuzi Zhou, Yankai Li and Mo Chen. (1/7)

Ke Li 🍁

102,877 görüntüleme • 5 ay önce

Generative models (diffusion/flow) are taking over robotics 🤖. But do we really need to model the full action distribution to control a robot? We suspected the success of Generative Control Policies (GCPs) might be "Much Ado About Noising." We rigorously tested the myths. 🧵👇

Generative models (diffusion/flow) are taking over robotics 🤖. But do we really need to model the full action distribution to control a robot? We suspected the success of Generative Control Policies (GCPs) might be "Much Ado About Noising." We rigorously tested the myths. 🧵👇

Chaoyi Pan

112,956 görüntüleme • 11 ay önce

How can we train generative models on 100 examples without pretraining on large datasets? Presenting Adaptive IMLE, which outperforms both GANs and diffusion models. Details at Work w/ Mehran Aghabozorgi and Shichong Peng. Come by our #ICML2023 poster Wed AM 1/2

How can we train generative models on 100 examples without pretraining on large datasets? Presenting Adaptive IMLE, which outperforms both GANs and diffusion models. Details at Work w/ Mehran Aghabozorgi and Shichong Peng. Come by our #ICML2023 poster Wed AM 1/2

Ke Li 🍁

15,388 görüntüleme • 3 yıl önce

Diffusion models make great images. But can they drive robots? Usually that gets complicated really fast. We figured out how to get a Stable Diffusion model (based on Instruct pix2pix) to drive robotic instruction following. Simple recipe, works on a wide range of tasks. Thread👇

Diffusion models make great images. But can they drive robots? Usually that gets complicated really fast. We figured out how to get a Stable Diffusion model (based on Instruct pix2pix) to drive robotic instruction following. Simple recipe, works on a wide range of tasks. Thread👇

Sergey Levine

126,572 görüntüleme • 3 yıl önce

[1/N] We present a plug-and-play mechanism to controllably steer inference of any diffusion/flow model towards a sharper or flatter sampling distribution, resulting in improvements across domains e.g. text-to-image (10% FID reduction), protein generation (improved designability).

[1/N] We present a plug-and-play mechanism to controllably steer inference of any diffusion/flow model towards a sharper or flatter sampling distribution, resulting in improvements across domains e.g. text-to-image (10% FID reduction), protein generation (improved designability).

Shubham Tulsiani

83,576 görüntüleme • 1 yıl önce

We just brought flow maps to language modeling for one-step sequence generation 💥 Discrete diffusion is not necessary -- continuous flows over one-hot encodings achieve SoTA performance and ≥8.3× faster generation 🔥 We believe this is a major step forward for discrete generative modeling and language modeling alike. 🚀 Full thread from first author Chanhyuk Lee:

We just brought flow maps to language modeling for one-step sequence generation 💥 Discrete diffusion is not necessary -- continuous flows over one-hot encodings achieve SoTA performance and ≥8.3× faster generation 🔥 We believe this is a major step forward for discrete generative modeling and language modeling alike. 🚀 Full thread from first author Chanhyuk Lee:

Nicholas Boffi

44,715 görüntüleme • 8 ay önce

We are releasing 4M-21 with a permissive license, including its source code and trained models. It's a pretty effective multimodal model that solves 10s of tasks & modalities. See the demo code, sample results, and the tokenizers of diverse modalities on the website. IMO, the multitask learning aspect of multimodal models has really taken a step forward. We can train a single model on many diverse tasks with ~SOTA accuracy. But a long way to go in terms of transfer/emergence. 🌐 ⌨️ Joint work w/ EPFL Apple.

We are releasing 4M-21 with a permissive license, including its source code and trained models. It's a pretty effective multimodal model that solves 10s of tasks & modalities. See the demo code, sample results, and the tokenizers of diverse modalities on the website. IMO, the multitask learning aspect of multimodal models has really taken a step forward. We can train a single model on many diverse tasks with ~SOTA accuracy. But a long way to go in terms of transfer/emergence. 🌐 ⌨️ Joint work w/ EPFL Apple.

Amir Zamir

69,627 görüntüleme • 2 yıl önce

VLA policies are slow and jerky. Introducing IMLE-VLA, which swaps out flow matching for conditional IMLE. IMLE-VLA generates action chunks at 55 Hz (3.67x faster than π0.5), produces >2.2x less jerk and finishes tasks >3.9x faster. Details at If you are at #IROS2026, don't miss Kian Hosseinkhani's talk on Monday at 9am in Room 329. Joint work w/ Kian Hosseinkhani, Qinhe Peng, George Shramko, Mehran Aghabozorgi, Aurora Qian, Tristan Engst, Alireza Moazeni and Dinesh Jayaraman Thread 👇 1/8

VLA policies are slow and jerky. Introducing IMLE-VLA, which swaps out flow matching for conditional IMLE. IMLE-VLA generates action chunks at 55 Hz (3.67x faster than π0.5), produces >2.2x less jerk and finishes tasks >3.9x faster. Details at If you are at #IROS2026, don't miss Kian Hosseinkhani's talk on Monday at 9am in Room 329. Joint work w/ Kian Hosseinkhani, Qinhe Peng, George Shramko, Mehran Aghabozorgi, Aurora Qian, Tristan Engst, Alireza Moazeni and Dinesh Jayaraman Thread 👇 1/8

Ke Li 🍁

12,771 görüntüleme • 1 ay önce

The co-founders of Flapping Airplanes call the current RL paradigm for model training "environment slop." They explain: "The reinforcement paradigms of today are shockingly inefficient. You don't really get much generalization across tasks, you teach a model through one kind of learning and then you teach it the next one. It's kind of like whack-a-mole. We look at this and think it's kind of crazy. The next paradigm of AI will not be environment slop." "Human level intelligence is not the ceiling, it is merely the floor on what is possible. If you can train models with vastly less data and possibly more compute in very different ways, what is going to happen? We actually don't know. But I do think they'll be different and weird and they'll have interesting capabilities that we'll find really valuable ways to use."

The co-founders of Flapping Airplanes call the current RL paradigm for model training "environment slop." They explain: "The reinforcement paradigms of today are shockingly inefficient. You don't really get much generalization across tasks, you teach a model through one kind of learning and then you teach it the next one. It's kind of like whack-a-mole. We look at this and think it's kind of crazy. The next paradigm of AI will not be environment slop." "Human level intelligence is not the ceiling, it is merely the floor on what is possible. If you can train models with vastly less data and possibly more compute in very different ways, what is going to happen? We actually don't know. But I do think they'll be different and weird and they'll have interesting capabilities that we'll find really valuable ways to use."

TBPN

85,358 görüntüleme • 9 ay önce

GenCast is a diffusion model, similar to the machine learning models which also power generative AI. 🎨 We trained it on 40 years of historical data from ECMWF, which included variables such as temperature, wind speed, and pressure at various altitudes - enabling it to learn global weather patterns.

GenCast is a diffusion model, similar to the machine learning models which also power generative AI. 🎨 We trained it on 40 years of historical data from ECMWF, which included variables such as temperature, wind speed, and pressure at various altitudes - enabling it to learn global weather patterns.

Google DeepMind

26,360 görüntüleme • 1 yıl önce

Excited to share our interview with Sander Dieleman! This is Part 3 in our History of Diffusion series. We talk about diffusion as spectral autoregression, diffusion language models, flow matching, and much more. Enjoy!

Excited to share our interview with Sander Dieleman! This is Part 3 in our History of Diffusion series. We talk about diffusion as spectral autoregression, diffusion language models, flow matching, and much more. Enjoy!

Slater Stich

28,721 görüntüleme • 1 yıl önce

We release a technical blog on training PocketTTS, our 100M parameter on-device TTS, through drifting, a recent one-step generative objective from Deng et al. Less than 1% WER, high quality and voice cloning. To the best of our knowledge, it's the first speech model, and the first autoregressive model, trained this way.

We release a technical blog on training PocketTTS, our 100M parameter on-device TTS, through drifting, a recent one-step generative objective from Deng et al. Less than 1% WER, high quality and voice cloning. To the best of our knowledge, it's the first speech model, and the first autoregressive model, trained this way.

kyutai

24,531 görüntüleme • 1 ay önce

“Soccer is not a men's game. It's the people's game. And we all belong. Young girls need to know….they belong on television for the entire world to see." This Sinclair speech is one of a kind. She is a role model, an icon, and a Portland legend.

“Soccer is not a men's game. It's the people's game. And we all belong. Young girls need to know….they belong on television for the entire world to see." This Sinclair speech is one of a kind. She is a role model, an icon, and a Portland legend.

SHE scores bangers

35,913 görüntüleme • 2 yıl önce

Generative #AI is creating new opportunities fast, but like-minded democracies need to get in control of the risks. As #G7 we took a big step forward today. We agreed draft guiding principles on generative AI for stakeholder consultation. We will 🔜 reach out in 🇪🇺! Stay tuned!

Generative #AI is creating new opportunities fast, but like-minded democracies need to get in control of the risks. As #G7 we took a big step forward today. We agreed draft guiding principles on generative AI for stakeholder consultation. We will 🔜 reach out in 🇪🇺! Stay tuned!

Věra Jourová

15,641 görüntüleme • 3 yıl önce

We present Neural Riemannian Motion Fields (NRMF) as a promising paradigm for generative modeling of articulated motion. Unlike various diffusion based models, NRMF is useful both in generation and in deployment as a prior across tasks: #CVPR #AI #CVML👇

We present Neural Riemannian Motion Fields (NRMF) as a promising paradigm for generative modeling of articulated motion. Unlike various diffusion based models, NRMF is useful both in generation and in deployment as a prior across tasks: #CVPR #AI #CVML👇

Tolga Birdal

16,795 görüntüleme • 1 yıl önce

One of the strengths of flow matching is that it decouples generative modeling from a specific noise process and choice of source distribution. You can select from a variety of initial distributions like a Gaussian or a uniform distribution and still train a valid flow model.

One of the strengths of flow matching is that it decouples generative modeling from a specific noise process and choice of source distribution. You can select from a variety of initial distributions like a Gaussian or a uniform distribution and still train a valid flow model.

Alec Helbling

48,705 görüntüleme • 9 ay önce

My argument on reconciliation is simple: Why go small, when you can go big? We need to make a real impact before November, and we should get that done now.

My argument on reconciliation is simple: Why go small, when you can go big? We need to make a real impact before November, and we should get that done now.

Senator Ted Cruz

51,930 görüntüleme • 6 ay önce

The best founders in the world like Elon Musk or tobi lutke “go ultra-hardcore on deletion and simplification.” Luca Ferrari says Bending Spoons employs this idea of “Relentless Simplification” across their businesses: “Humans have a tendency to add complexity and do things that destroy value.” “If you leave an organization, almost any environment unattended, and you don't provide guidance in this regard, it will tend to become more complicated. People will add parts. It could be expanding a team or adding a step to a process, or adding an entire new process. If it's a product, adding a feature to the product. New rules. It applies to almost any human endeavor.” “People tend to add pieces, they very rarely remove pieces.” “If you don't make a conscious effort to achieve simplicity, and avoid this increasing value-destroying complexity, that's how we got to our modern society with all the bureaucracy and complicated regulation.” “We have this principle whereby we ask everyone who works here, every time someone is suggesting that we should be adding complexity, the burden of proof is on those making that suggestion. The people who support the thesis that we shouldn't be adding complexity don't need to prove it. They just have to raise a flag and say, ‘I don't think we should.’ The burden of proof is on those who want to add complexity.” “The other part of relentless simplification is that we want people to be on the lookout for existing complexity and suggest that we should remove it.”

The best founders in the world like Elon Musk or tobi lutke “go ultra-hardcore on deletion and simplification.” Luca Ferrari says Bending Spoons employs this idea of “Relentless Simplification” across their businesses: “Humans have a tendency to add complexity and do things that destroy value.” “If you leave an organization, almost any environment unattended, and you don't provide guidance in this regard, it will tend to become more complicated. People will add parts. It could be expanding a team or adding a step to a process, or adding an entire new process. If it's a product, adding a feature to the product. New rules. It applies to almost any human endeavor.” “People tend to add pieces, they very rarely remove pieces.” “If you don't make a conscious effort to achieve simplicity, and avoid this increasing value-destroying complexity, that's how we got to our modern society with all the bureaucracy and complicated regulation.” “We have this principle whereby we ask everyone who works here, every time someone is suggesting that we should be adding complexity, the burden of proof is on those making that suggestion. The people who support the thesis that we shouldn't be adding complexity don't need to prove it. They just have to raise a flag and say, ‘I don't think we should.’ The burden of proof is on those who want to add complexity.” “The other part of relentless simplification is that we want people to be on the lookout for existing complexity and suggest that we should remove it.”

David Senra

309,130 görüntüleme • 1 ay önce

People keep talking about HertzFlow | Mainnet Arc🎲's RFQ model. But there is one question that deserves an answer and of which is simple: Why use an RFQ instead of the normal order book most traders already understand? And to answer this question, we first need to understand what an order book actually does. #HertzFlow A thread 🧵 and a Video 🎥

People keep talking about HertzFlow | Mainnet Arc🎲's RFQ model. But there is one question that deserves an answer and of which is simple: Why use an RFQ instead of the normal order book most traders already understand? And to answer this question, we first need to understand what an order book actually does. #HertzFlow A thread 🧵 and a Video 🎥

Cleopatra

18,790 görüntüleme • 26 gün önce