Loading video...

Video Failed to Load

There was a problem loading this video. This could be due to a temporary network issue or the video might be unavailable.

📢 New paper and 🤗 demo! Can LLM systems solve the classic program synthesis problem - generate code from example test cases alone? We use 🌱 Synthetic data generation from seed problems and finetuning 📈 Test time compute - draw more samples and filter! (1/n)

Wen-Ding Li

3,093 subscribers

16,305 views • 2 years ago •via X (Twitter)

Science & Technology News & Politics Education

Anya Rossi• Live Now

Private livecam show

0 Comments

No comments available

Comments from the original post will appear here

Related Videos

We’ve added interactive visualizations to all coding problems. You can now understand each coding problem more clearly and in depth before solving. Here’s an example using positional encodings from the transformer paper.

We’ve added interactive visualizations to all coding problems. You can now understand each coding problem more clearly and in depth before solving. Here’s an example using positional encodings from the transformer paper.

TensorTonic

33,033 views • 5 months ago

I created so many design systems this year and kept creating the same tokens over and over. So, I’m creating this tiny Figma plugin to solve the 'blank canvas' problem. Here small test demo to create color and text tokens 🛠️

I created so many design systems this year and kept creating the same tokens over and over. So, I’m creating this tiny Figma plugin to solve the 'blank canvas' problem. Here small test demo to create color and text tokens 🛠️

vijay verma

49,588 views • 1 year ago

New RFC - Override HTTP response headers in Chrome DevTools. Here's a demo on how to override and test CORS headers. It support more use-cases than that! Give the prototype a spin and share your thoughts 👉🏼

New RFC - Override HTTP response headers in Chrome DevTools. Here's a demo on how to override and test CORS headers. It support more use-cases than that! Give the prototype a spin and share your thoughts 👉🏼

Jecelyn Yeen 🐟

39,239 views • 3 years ago

Seedance 2.0 Test Demo 03 Use storyboard picture as the input and generate a Y2K pop style MV scene Great work Dreamina! Dreamina AI

Seedance 2.0 Test Demo 03 Use storyboard picture as the input and generate a Y2K pop style MV scene Great work Dreamina! Dreamina AI

Jim HuiHui

71,850 views • 4 months ago

How can we use test-time compute for spatial understanding? 🤔 In InterPose, we propose to repeatedly sample generative video models to help two-view pose estimation and reconstruction, by leveraging the video models' keyframe interpolation abilities. A 🧵... (1/8)

How can we use test-time compute for spatial understanding? 🤔 In InterPose, we propose to repeatedly sample generative video models to help two-view pose estimation and reconstruction, by leveraging the video models' keyframe interpolation abilities. A 🧵... (1/8)

Ricardo Martin-Brualla

21,685 views • 1 year ago

Our new Mistral AI playground is now live! 🔥 You can now quickly test our models in the playground and switch easily from UX to code. Try it out and more exciting features coming soon!

Our new Mistral AI playground is now live! 🔥 You can now quickly test our models in the playground and switch easily from UX to code. Try it out and more exciting features coming soon!

Sophia Yang, Ph.D.

19,557 views • 1 year ago

Introducing crypto on Dexter We can now: • track prices • research ideas • test correlations Across hundreds of cryptocurrencies. Data from Coinbase, Kraken, and Bitfinex.

Introducing crypto on Dexter We can now: • track prices • research ideas • test correlations Across hundreds of cryptocurrencies. Data from Coinbase, Kraken, and Bitfinex.

Virat Singh

24,395 views • 6 months ago

(1/n) 🚀 With FastVideo, you can now generate a 5-second video in 5 seconds on a single H200 GPU! Introducing FastWan series, a family of fast video generation models trained via a new recipe we term as “sparse distillation”, to speed up video denoising time by 70X! 🖥️ Live demo: (Thanks to @gmicloud for the support!) 🔗 Blog: 🔓 We fully open-source our models, code, and data with Apache-2.0 licenses

(1/n) 🚀 With FastVideo, you can now generate a 5-second video in 5 seconds on a single H200 GPU! Introducing FastWan series, a family of fast video generation models trained via a new recipe we term as “sparse distillation”, to speed up video denoising time by 70X! 🖥️ Live demo: (Thanks to @gmicloud for the support!) 🔗 Blog: 🔓 We fully open-source our models, code, and data with Apache-2.0 licenses

Hao AI Lab

78,660 views • 10 months ago

Starship's eleventh flight test reached every objective, providing valuable data as we prepare the next generation of Starship and Super Heavy →

Starship's eleventh flight test reached every objective, providing valuable data as we prepare the next generation of Starship and Super Heavy →

SpaceX

870,205 views • 8 months ago

Get free access to Supermicro’s top-end data center servers just from your browser. Our JumpStart program offers remote access to our top-end H13 and H14 systems for qualified customers to validate, test, and benchmark their workloads with simple steps.

Get free access to Supermicro’s top-end data center servers just from your browser. Our JumpStart program offers remote access to our top-end H13 and H14 systems for qualified customers to validate, test, and benchmark their workloads with simple steps.

Supermicro

9,796,966 views • 8 months ago

Turn complex docs into clean, LLM-ready data! Every AI company I've talked to is solving the same problem: how do you build systems that don't hallucinate and back up every answer with proper citations? Tensorlake is a tool that extracts custom-defined structured data from any unstructured document in 3 steps: ↳ Define your schema ↳ Enable citations ↳ Extract You get RAG-ready data with precise citations and bounding boxes. Feed this to your LLM, and you'll generate responses that are citation-backed and fully auditable. This is the difference between a demo and a production system. When your AI can show exactly where it got its information, you move from proof-of-concept to something people can actually trust and deploy. I've shared the Tensorlake GitHub repo in the replies!

Turn complex docs into clean, LLM-ready data! Every AI company I've talked to is solving the same problem: how do you build systems that don't hallucinate and back up every answer with proper citations? Tensorlake is a tool that extracts custom-defined structured data from any unstructured document in 3 steps: ↳ Define your schema ↳ Enable citations ↳ Extract You get RAG-ready data with precise citations and bounding boxes. Feed this to your LLM, and you'll generate responses that are citation-backed and fully auditable. This is the difference between a demo and a production system. When your AI can show exactly where it got its information, you move from proof-of-concept to something people can actually trust and deploy. I've shared the Tensorlake GitHub repo in the replies!

Akshay 🚀

58,117 views • 7 months ago

My first PhD paper!🎉We learn *diffusion* models for code generation that learn to directly *edit* syntax trees of programs. The result is a system that can incrementally write code, see the execution output, and debug it. 🧵1/n

My first PhD paper!🎉We learn diffusion models for code generation that learn to directly edit syntax trees of programs. The result is a system that can incrementally write code, see the execution output, and debug it. 🧵1/n

Shreyas Kapur

742,465 views • 2 years ago

How can we address the scarcity of data required for specialized AI? Learn about Simula, a framework that reframes synthetic data generation as dataset-level mechanism design. By using reasoning to architect datasets from first principles, Simula enables fine-grained control over coverage, complexity, and quality. More →

How can we address the scarcity of data required for specialized AI? Learn about Simula, a framework that reframes synthetic data generation as dataset-level mechanism design. By using reasoning to architect datasets from first principles, Simula enables fine-grained control over coverage, complexity, and quality. More →

Google Research

142,219 views • 2 months ago

The use cases for blockchain adoption are endless… at least when it comes to DevvX. Think beyond the usual: identity verification, voting systems, notary services, copyright protection, and more. We can plug into existing web2 structures easily through RESTful API and solve real world challenges for businesses.😏

The use cases for blockchain adoption are endless… at least when it comes to DevvX. Think beyond the usual: identity verification, voting systems, notary services, copyright protection, and more. We can plug into existing web2 structures easily through RESTful API and solve real world challenges for businesses.😏

DevvE

15,332 views • 1 year ago

Some BTS from our durability labs to show how #Pixel10 and Pixel 10 Pro withstand the test of time—and the test of drops, spills, scratches, and oopsies in between¹ 🙌 BTW, each Pixel passes a 1 meter granite drop test 🪨 Go deep on durability:

Some BTS from our durability labs to show how #Pixel10 and Pixel 10 Pro withstand the test of time—and the test of drops, spills, scratches, and oopsies in between¹ 🙌 BTW, each Pixel passes a 1 meter granite drop test 🪨 Go deep on durability:

Made by Google

24,694 views • 5 months ago

We have released the code and weights for our #CVPR2023 paper "Avatars Grow Legs: Generating Smooth Human Motion from Sparse Tracking Inputs with Diffusion Model"! code: abs: project: The demo is below:

We have released the code and weights for our #CVPR2023 paper "Avatars Grow Legs: Generating Smooth Human Motion from Sparse Tracking Inputs with Diffusion Model"! code: abs: project: The demo is below:

Artsiom Sanakoyeu

35,711 views • 3 years ago

Announcing "look at your damn data" for Reinforcement Learning! You now get the full power of W&B Weave traces in the wan deeee bee workspace. For any wandb run that calls an LLM, just "pip install weave" and "import weave" in your code. We also have out-of-the-box integrations for verifiers from will brown, ART from OpenPipe, and verl from Bytedance Seed Team. Use those libraries with wandb/weave installed and everything just works.

Announcing "look at your damn data" for Reinforcement Learning! You now get the full power of W&B Weave traces in the wan deeee bee workspace. For any wandb run that calls an LLM, just "pip install weave" and "import weave" in your code. We also have out-of-the-box integrations for verifiers from will brown, ART from OpenPipe, and verl from Bytedance Seed Team. Use those libraries with wandb/weave installed and everything just works.

Shawn Lewis

37,581 views • 9 months ago

Our CUDA-native genome generation stack can now generate a whole synthetic genome in just under 20 minutes on eight H200 GPUs from lium.io . That's roughly 440x faster than the traditional CPU-bound workflows. This opens the door to population-scale synthetic genome generation for benchmarking, variant calling evaluation, and genomic AI.

Our CUDA-native genome generation stack can now generate a whole synthetic genome in just under 20 minutes on eight H200 GPUs from lium.io . That's roughly 440x faster than the traditional CPU-bound workflows. This opens the door to population-scale synthetic genome generation for benchmarking, variant calling evaluation, and genomic AI.

Minos

15,421 views • 10 days ago

Robotics has a dirty secret. The biggest bottleneck isn't hardware. It isn't compute. It's data. And right now, only a handful of corporations can afford to generate it... $ROBA No Code Studio is about to change that. 🧵(1/8)

Robotics has a dirty secret. The biggest bottleneck isn't hardware. It isn't compute. It's data. And right now, only a handful of corporations can afford to generate it... $ROBA No Code Studio is about to change that. 🧵(1/8)

Roba Labs

23,621 views • 1 month ago

Testing something new for Ronin Market. The pace of innovation coming from SM is accelerating. Time to test, build, and set new metas 🛠️

Testing something new for Ronin Market. The pace of innovation coming from SM is accelerating. Time to test, build, and set new metas 🛠️

Jihoz.ron

15,383 views • 1 year ago