Video yükleniyor...

Video Yüklenemedi

Ana Sayfaya Dön

We released invent a dataset. Today we release the technical report that goes with it. Huge shoutout to everyone Shivalika Singh Andrija Djurisic Leke Onilude Sudip Roy 🔥 This tackles the hardest setting. How do you go from zero data regime to a training set for a capability.

sarahookr's profile picture

Sara Hooker

64,947 subscribers

41,860 görüntüleme • 1 ay önce •via X (Twitter)

Bilim & Teknoloji
LIVE

Anya Rossi• Live Now

Private livecam show

18 Yorum

Sara Hooker profil fotoğrafı
Sara Hooker1 ay önce

Full technical report @askalphaxiv: 🔥🔥 You can just start inventing here:

Tim Davidson profil fotoğrafı
Tim Davidson1 ay önce

@singhshiviii @andrijazzz @lekeonilude @sudip_r0y Hi Sara, you might want to check out our TMLR work on seedless synthetic data generation: Also featured here:

Sara Hooker profil fotoğrafı
Sara Hooker1 ay önce

@singhshiviii @andrijazzz @lekeonilude @sudip_r0y Happy to take a look, and will make sure we reference if we don’t already do so. Seems extremely relevant.

Tim Davidson profil fotoğrafı
Tim Davidson1 ay önce

@singhshiviii @andrijazzz @lekeonilude @sudip_r0y 🙏 — also; congrats on the launch!

Connor Shorten profil fotoğrafı
Connor Shorten1 ay önce

@singhshiviii @andrijazzz @lekeonilude @sudip_r0y 🔥🔥

Ahmad Mustafa Anis profil fotoğrafı
Ahmad Mustafa Anis1 ay önce

@singhshiviii @andrijazzz @lekeonilude @sudip_r0y very cool

goubie profil fotoğrafı
goubie1 ay önce

@singhshiviii @andrijazzz @lekeonilude @sudip_r0y I used it twice already: a São Paulo-specific conversation dataset becoming a Gemma finetuning, and one specific to Brazilian innovation laws which still needs to be used in finetuning. Recommended!

Sara Hooker profil fotoğrafı
Sara Hooker1 ay önce

@singhshiviii @andrijazzz @lekeonilude @sudip_r0y 🔥

Sophia Yang profil fotoğrafı
Sophia Yang1 ay önce

@singhshiviii @andrijazzz @lekeonilude @sudip_r0y Congrats 🎉

Sara Hooker profil fotoğrafı
Sara Hooker1 ay önce

@singhshiviii @andrijazzz @lekeonilude @sudip_r0y Thank you Sophia! Congrats to you too on your new chapter. 🔥

Sophia Yang profil fotoğrafı
Sophia Yang1 ay önce

@singhshiviii @andrijazzz @lekeonilude @sudip_r0y 🫶🫶🥰

Eliza Kosoy profil fotoğrafı
Eliza Kosoy1 ay önce

@singhshiviii @andrijazzz @lekeonilude @sudip_r0y speechless at how fast this generates 😭😭😭

Nahid Alam profil fotoğrafı
Nahid Alam1 ay önce

@singhshiviii @andrijazzz @lekeonilude @sudip_r0y cool, congrats on the launch!

Assaf Petronio profil fotoğrafı
Assaf Petronio1 ay önce

@singhshiviii @andrijazzz @lekeonilude @sudip_r0y cool direction. curious how you check coverage when there are no real examples to compare against

Jatin Garg profil fotoğrafı
Jatin Garg1 ay önce

@singhshiviii @andrijazzz @lekeonilude @sudip_r0y synthetic data quality is the hard part. how did you validate the invented examples before using them for training?

玻璃渣 profil fotoğrafı
玻璃渣1 ay önce

@singhshiviii @andrijazzz @lekeonilude @sudip_r0y Data being the biggest bottleneck feels underrated, curious to dig into this.

ossama profil fotoğrafı
ossama1 ay önce

@singhshiviii @andrijazzz @lekeonilude @sudip_r0y @grok is this thing opensource?

Vedant Korade profil fotoğrafı
Vedant Korade1 ay önce

@singhshiviii @andrijazzz @lekeonilude @sudip_r0y it's basically a llm agent with a harness accessed via a paid api endpoint

Benzer Videolar

Last week we introduced Invent a Dataset. Describe what you need. Hit go. Today, we’re making it even easier with the Invent API. A few lines of code. AI ready training datasets in minutes.
0:23

Last week we introduced Invent a Dataset. Describe what you need. Hit go. Today, we’re making it even easier with the Invent API. A few lines of code. AI ready training datasets in minutes.

adaption

61,678 görüntüleme • 1 ay önce

Phang: This is a win that we must all celebrate together because this is how we do it. We find creative ways to think outside the box to force this regime to actually obey the law, the laws that you and I and everyone else would be forced to do.
0:28

Phang: This is a win that we must all celebrate together because this is how we do it. We find creative ways to think outside the box to force this regime to actually obey the law, the laws that you and I and everyone else would be forced to do.

MeidasClips

27,901 görüntüleme • 4 ay önce

Very excited to release the Open X-Embodiment Dataset today — the largest robot dataset to date with 1M+ trajectories! Robotics needs more data & this is a big step! There’s lots to unpack here, so let’s do a deep dive into the dataset! 🧵1/15
0:36

Very excited to release the Open X-Embodiment Dataset today — the largest robot dataset to date with 1M+ trajectories! Robotics needs more data & this is a big step! There’s lots to unpack here, so let’s do a deep dive into the dataset! 🧵1/15

Karl Pertsch

111,195 görüntüleme • 3 yıl önce

I’m delighted that we could contribute the #cryoEM data to this incredible antiviral discovery project from the group of Johan Neyts 💊 Published today in Nature: A huge congrats to everyone involved! 🎉
0:22

I’m delighted that we could contribute the #cryoEM data to this incredible antiviral discovery project from the group of Johan Neyts 💊 Published today in Nature: A huge congrats to everyone involved! 🎉

Daniel Hurdiss

21,379 görüntüleme • 1 yıl önce

Britain was built for winters - so how does it survive extreme heat? With temperatures scorching the country this week, we need to change how we live during a heatwave, according to the latest report from the Climate Change Committee. From setting maximum temperatures for workers, to installing air con in care homes - the advice states that the UK needs to adapt.
1:37

Britain was built for winters - so how does it survive extreme heat? With temperatures scorching the country this week, we need to change how we live during a heatwave, according to the latest report from the Climate Change Committee. From setting maximum temperatures for workers, to installing air con in care homes - the advice states that the UK needs to adapt.

Channel 4 News

23,867 görüntüleme • 5 ay önce

The internet is the greatest dataset ever created. Today, we're launching the Riveter Dataset Builder to make it possible for anyone to get custom, fresh data from a prompt.
0:33

The internet is the greatest dataset ever created. Today, we're launching the Riveter Dataset Builder to make it possible for anyone to get custom, fresh data from a prompt.

Abby Grills

24,870 görüntüleme • 6 ay önce

Today we are glad to present our new project in collaboration with Fabio Montresor! WORLD CHAMPIONSHIP TRAINING! A pioneer project that will reshape the way we want to share our knowledge about One piece with everyone! In this series you will find the whole process we go through to prepare a tournament! Link to the project: We will release a new episode every week included for free for all the owners! Check out the first part for free on Youtube! Link below 👇👇👇
1:00

Today we are glad to present our new project in collaboration with Fabio Montresor! WORLD CHAMPIONSHIP TRAINING! A pioneer project that will reshape the way we want to share our knowledge about One piece with everyone! In this series you will find the whole process we go through to prepare a tournament! Link to the project: We will release a new episode every week included for free for all the owners! Check out the first part for free on Youtube! Link below 👇👇👇

Dog of Wisdom

10,825 görüntüleme • 8 ay önce

Today, we release Futardio. The place for founders who want to go faster. Forget pitching VCs - raise directly from a community that wants you to win from day #1.
0:49

Today, we release Futardio. The place for founders who want to go faster. Forget pitching VCs - raise directly from a community that wants you to win from day #1.

MetaDAO

250,277 görüntüleme • 8 ay önce

Introducing ABC: open data, training, and infrastructure for robotics. We release the largest teleop dataset to date, and extensively investigate design decisions, pretraining, and post-training techniques. Arthur Allshire Himanshu Gaurav Singh Adam Rashid Hongsuk Benjamin Choi David McAllister
2:51

Introducing ABC: open data, training, and infrastructure for robotics. We release the largest teleop dataset to date, and extensively investigate design decisions, pretraining, and post-training techniques. Arthur Allshire Himanshu Gaurav Singh Adam Rashid Hongsuk Benjamin Choi David McAllister

Ritvik Singh

372,221 görüntüleme • 4 ay önce

This is a historic data release. A real advance for pancreatic cancer patients, and it shows that our industry goes after the hardest things and can win. #ASCO26
7:21

This is a historic data release. A real advance for pancreatic cancer patients, and it shows that our industry goes after the hardest things and can win. #ASCO26

Brad Loncar

26,052 görüntüleme • 5 ay önce

One of the cooler under the hood insights I’ve heard from a head coach at a postgame presser. We talk about the data teams have access to that we don’t. This is a great example of the level of detail that goes into setting a rotation and allocating minutes.
1:10

One of the cooler under the hood insights I’ve heard from a head coach at a postgame presser. We talk about the data teams have access to that we don’t. This is a great example of the level of detail that goes into setting a rotation and allocating minutes.

Rit Holtzman

207,686 görüntüleme • 23 saat önce

How do you scale 2 million hours of ego data without sacrificing quality? Earlier this week, we revealed the world’s largest egocentric dataset. Our frontier pipeline has also turned it into the most labeled, annotated, and human QA’ed robotics ego dataset in the world. We’ve focused on some of the hardest problems in computer vision and ego data to ensure every clip contains SOTA action supervision directly helpful for training dexterous manipulation.
0:21

How do you scale 2 million hours of ego data without sacrificing quality? Earlier this week, we revealed the world’s largest egocentric dataset. Our frontier pipeline has also turned it into the most labeled, annotated, and human QA’ed robotics ego dataset in the world. We’ve focused on some of the hardest problems in computer vision and ego data to ensure every clip contains SOTA action supervision directly helpful for training dexterous manipulation.

Midcentury

10,802 görüntüleme • 1 ay önce

We release Action100M, the hero behind VL-JEPA. It is a large dataset with O(100 million) dense action annotations on HowTo100M procedural videos. We hope it serves as a robust data foundation to advance physical world modeling research.
0:40

We release Action100M, the hero behind VL-JEPA. It is a large dataset with O(100 million) dense action annotations on HowTo100M procedural videos. We hope it serves as a robust data foundation to advance physical world modeling research.

Delong Chen (陈德龙)

103,886 görüntüleme • 9 ay önce

LOP Joel Ssenyonyi to Speaker Annet Anitah Among : “The committee report to this bill (Protection of Sovereignty Bill,2026) leaked . On 18th January 2023, I rose to present a committee report for COSASE on Uganda Airlines , I want to quote you verbatim what you said on the hansard. There was a leakage of the Uganda Airlines report , a property of this Parliament cannot leak and we continue debating on it.’’ Rt.Hon Speaker this report has leaked , so given that precedent that you set ,are we going to go ahead and debate this leaked report. “ ?
1:11

LOP Joel Ssenyonyi to Speaker Annet Anitah Among : “The committee report to this bill (Protection of Sovereignty Bill,2026) leaked . On 18th January 2023, I rose to present a committee report for COSASE on Uganda Airlines , I want to quote you verbatim what you said on the hansard. There was a leakage of the Uganda Airlines report , a property of this Parliament cannot leak and we continue debating on it.’’ Rt.Hon Speaker this report has leaked , so given that precedent that you set ,are we going to go ahead and debate this leaked report. “ ?

Gideon Nova Kwikiriza

62,230 görüntüleme • 6 ay önce

1 day to go! Our first episode of #BuildTheBlock goes live tomorrow, but in the meantime, we thought we'd give you a behind-the-scenes look at how we made it all possible. Set a reminder for Episode 1 👇
1:00

1 day to go! Our first episode of #BuildTheBlock goes live tomorrow, but in the meantime, we thought we'd give you a behind-the-scenes look at how we made it all possible. Set a reminder for Episode 1 👇

Binance

334,281 görüntüleme • 3 yıl önce

"One of the very confusing things about the models right now: how to reconcile the fact that they are doing so well on evals. And you look at the evals and you go, 'Those are pretty hard evals.' But the economic impact seems to be dramatically behind. There is [a possible] explanation. Back when people were doing pre-training, the question of what data to train on was answered, because that answer was everything. So you don't have to think if it's going to be this data or that data. When people do RL training, they say, 'Okay, we want to have this kind of RL training for this thing and that kind of RL training for that thing.' You say, 'Hey, I would love our model to do really well when we release it. I want the evals to look great. What would be RL training that could help on this task?' If you combine this with generalization of the models actually being inadequate, that has the potential to explain a lot of what we are seeing, this disconnect between eval performance and actual real-world performance"
2:21

"One of the very confusing things about the models right now: how to reconcile the fact that they are doing so well on evals. And you look at the evals and you go, 'Those are pretty hard evals.' But the economic impact seems to be dramatically behind. There is [a possible] explanation. Back when people were doing pre-training, the question of what data to train on was answered, because that answer was everything. So you don't have to think if it's going to be this data or that data. When people do RL training, they say, 'Okay, we want to have this kind of RL training for this thing and that kind of RL training for that thing.' You say, 'Hey, I would love our model to do really well when we release it. I want the evals to look great. What would be RL training that could help on this task?' If you combine this with generalization of the models actually being inadequate, that has the potential to explain a lot of what we are seeing, this disconnect between eval performance and actual real-world performance"

Dwarkesh Patel

502,761 görüntüleme • 11 ay önce

“This is a disgrace to the medical profession.” RFK Jr. slams a reporter who questions HHS’s science behind banning gender-affirming care in minors. “We have a 1,300 page report, I urge you to read it.” “I don't think anybody can read that science without saying we made a huge mistake.” “It's a disgrace to the medical establishment that they let this go on for so long based upon so little positive data and overwhelming data against these procedures.”
0:31

“This is a disgrace to the medical profession.” RFK Jr. slams a reporter who questions HHS’s science behind banning gender-affirming care in minors. “We have a 1,300 page report, I urge you to read it.” “I don't think anybody can read that science without saying we made a huge mistake.” “It's a disgrace to the medical establishment that they let this go on for so long based upon so little positive data and overwhelming data against these procedures.”

End Tribalism in Politics

156,680 görüntüleme • 10 ay önce

🚨🚨🚨 We have released a database with the most searched keywords on the App Store, free for everyone! Just yesterday, we sent a message to all Astro ASO Tool users informing them that we had completed the integration of a database with one million keywords into our backend. This database allows us to access increasingly reliable data on popularity Today, we want to take a further step forward by making this database with 1 million keywords with a popularity value > 5 publicly available to everyone on our website! In this video, Alice Ercolani explains how to use this tool to discover new keywords to track in Astro! Link to the tool in the next post!
2:19

🚨🚨🚨 We have released a database with the most searched keywords on the App Store, free for everyone! Just yesterday, we sent a message to all Astro ASO Tool users informing them that we had completed the integration of a database with one million keywords into our backend. This database allows us to access increasingly reliable data on popularity Today, we want to take a further step forward by making this database with 1 million keywords with a popularity value > 5 publicly available to everyone on our website! In this video, Alice Ercolani explains how to use this tool to discover new keywords to track in Astro! Link to the tool in the next post!

Matteo Spada

53,597 görüntüleme • 1 yıl önce

Ladies and gentlemen, allow me to report that a GOOD time was had! 🙌🏿🔥 To everyone who missed Behind The Content this year… next year tuli Namwe! We are coming back bigger and better. A huge shoutout to A Pass , Pumla Nabachwa , Kansiime Anne Entertainer and iAm Uncle Mo
2:39

Ladies and gentlemen, allow me to report that a GOOD time was had! 🙌🏿🔥 To everyone who missed Behind The Content this year… next year tuli Namwe! We are coming back bigger and better. A huge shoutout to A Pass , Pumla Nabachwa , Kansiime Anne Entertainer and iAm Uncle Mo

The Ugandan Boy Talk Show

19,639 görüntüleme • 4 ay önce