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.

We are moving from model controller to RL for locomotion. The policy is starting to look quite robust. Even more convinced now that legs are easier than wheels, given the software complexity has shrunk tremendously with RL.

Sankaet

14,482 subscribers

133,130 views • 1 year ago •via X (Twitter)

Anya Rossi• Live Now

Private livecam show

0 Comments

No comments available

Comments from the original post will appear here

Related Videos

enjoyed hosting CNET at Foundation Robotics SF office. We discuss defense usecases, teleop, RL locomotion and more

enjoyed hosting CNET at Foundation Robotics SF office. We discuss defense usecases, teleop, RL locomotion and more

Sankaet

113,210 views • 11 months ago

You'd think the race to AGI would mean training the biggest possible model. But parameter scaling had stalled for a long time after GPT-4's trillion+ parameters, and only now are models getting bigger again. What gives? Partially it’s RL scaling, as Dylan Patel explains. A 5T parameter model takes 5x longer to generate RL rollouts than a 1T model. Even if the bigger model is 2x more sample-efficient, the smaller model finishes RL faster, gets deployed to research sooner, and starts helping build the next model before the big one is even done training.

You'd think the race to AGI would mean training the biggest possible model. But parameter scaling had stalled for a long time after GPT-4's trillion+ parameters, and only now are models getting bigger again. What gives? Partially it’s RL scaling, as Dylan Patel explains. A 5T parameter model takes 5x longer to generate RL rollouts than a 1T model. Even if the bigger model is 2x more sample-efficient, the smaller model finishes RL faster, gets deployed to research sooner, and starts helping build the next model before the big one is even done training.

Dwarkesh Patel

65,123 views • 5 months ago

Behavioral Foundation Models (BFMs) trained with RL are secretly more powerful than we think. BFM’s directly output a policy believed to be near-optimal given any reward function. Our new work shows that they can actually do much better:

Behavioral Foundation Models (BFMs) trained with RL are secretly more powerful than we think. BFM’s directly output a policy believed to be near-optimal given any reward function. Our new work shows that they can actually do much better:

Harshit Sikchi

44,475 views • 1 year ago

We developed an RL method for fine-tuning our models for precise tasks in just a few hours or even minutes. Instead of training the whole model, we add an “RL token” output to π-0.6, our latest model, which is used by a tiny actor and critic to learn quickly with RL.

We developed an RL method for fine-tuning our models for precise tasks in just a few hours or even minutes. Instead of training the whole model, we add an “RL token” output to π-0.6, our latest model, which is used by a tiny actor and critic to learn quickly with RL.

Physical Intelligence

443,654 views • 6 months ago

The best way to get robust, high-quality robot performance is through reinforcement learning; but RL in either the real world or a traditional simulation has lots of limitations. Instead, Jiazhi Yang in RISE does RL in a compositional world model. Learn more ->

The best way to get robust, high-quality robot performance is through reinforcement learning; but RL in either the real world or a traditional simulation has lots of limitations. Instead, Jiazhi Yang in RISE does RL in a compositional world model. Learn more ->

Chris Paxton

34,102 views • 3 months ago

Open Duck Mini is standing up using a policy learned with RL in simulation! We are still pushing locomotion, so if you have some knowledge of Isaac Gym and want to help feel free to join us! Thomas Wolf

Open Duck Mini is standing up using a policy learned with RL in simulation! We are still pushing locomotion, so if you have some knowledge of Isaac Gym and want to help feel free to join us! Thomas Wolf

Antoine Pirrone

66,116 views • 2 years ago

Hiring RL Engineer! Started off as a curious project at Lossfunk to push the boundaries of LLMs in social reasoning - we are now building RL environments, data, and benchmarks to simulate more real-world scenarios. If you want to train SoTA RL models over multi-GPUs (H200s/B200s) to unlock next AI frontier, this is for you.

Hiring RL Engineer! Started off as a curious project at Lossfunk to push the boundaries of LLMs in social reasoning - we are now building RL environments, data, and benchmarks to simulate more real-world scenarios. If you want to train SoTA RL models over multi-GPUs (H200s/B200s) to unlock next AI frontier, this is for you.

Satpal Singh Rathore

46,010 views • 1 year ago

We are slowly leaving the “AI chatbot” era. agents are starting to control actual software. watching Mercury Agent interact with Spotify natively from inside the workflow feels way more important than people realize right now.

We are slowly leaving the “AI chatbot” era. agents are starting to control actual software. watching Mercury Agent interact with Spotify natively from inside the workflow feels way more important than people realize right now.

GitHub Projects Community

27,280 views • 4 months ago

Is It All Starting To Make Sense Now? Lies Are Easier To Believe Than The Truth

Is It All Starting To Make Sense Now? Lies Are Easier To Believe Than The Truth

{Matt} $XRPatriot

364,466 views • 2 years ago

We asked Sholto Douglas from Anthropic about the costs of RL (Reinforcement Learning) runs. "In Dario Amodei's essay, he said that RL runs cost only $1M back in December." "RL is a more naively parallelizable and scalable than pre-training." "With pre-training, you need everything in one big data center ideally. For RL, in theory, you could scale all over the world."

We asked Sholto Douglas from Anthropic about the costs of RL (Reinforcement Learning) runs. "In Dario Amodei's essay, he said that RL runs cost only $1M back in December." "RL is a more naively parallelizable and scalable than pre-training." "With pre-training, you need everything in one big data center ideally. For RL, in theory, you could scale all over the world."

TBPN

76,696 views • 1 year ago

"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"

"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

502,484 views • 9 months ago

How should we pretrain a policy from demonstrations to ensure it is an effective initialization for RL finetuning, while preserving the performance of the pretrained policy itself? We propose Posterior Behavioral Cloning (PostBC)! (1/11)

How should we pretrain a policy from demonstrations to ensure it is an effective initialization for RL finetuning, while preserving the performance of the pretrained policy itself? We propose Posterior Behavioral Cloning (PostBC)! (1/11)

Andrew Wagenmaker

60,181 views • 9 months ago

I hand-wrote a 500-LoC RL stack to make hacking on RL research much easier. Most RL stacks are either massive and unhackable, or duct-taped research scripts. I am open-sourcing Mithrl, a modular RLVR stack. Next items on my checklist: adding more complex environment examples, supporting multi-gpu + async RL, and QoL fixes. I might scrap external runtime dependencies (Huggingface PEFT + vLLM) and write purpose-built, simpler versions from scratch if I feel the need. If you want to experiment with RL and are looking to own sovereign tools, I’d love to get on call, understand your requirements and help integrate for free.

I hand-wrote a 500-LoC RL stack to make hacking on RL research much easier. Most RL stacks are either massive and unhackable, or duct-taped research scripts. I am open-sourcing Mithrl, a modular RLVR stack. Next items on my checklist: adding more complex environment examples, supporting multi-gpu + async RL, and QoL fixes. I might scrap external runtime dependencies (Huggingface PEFT + vLLM) and write purpose-built, simpler versions from scratch if I feel the need. If you want to experiment with RL and are looking to own sovereign tools, I’d love to get on call, understand your requirements and help integrate for free.

omkaar

17,676 views • 6 months ago

Introducing INTELLECT-3: Scaling RL to a 100B+ MoE model on our end-to-end stack Achieving state-of-the-art performance for its size across math, code and reasoning Built using the same tools we put in your hands, from environments & evals, RL frameworks, sandboxes & more

Introducing INTELLECT-3: Scaling RL to a 100B+ MoE model on our end-to-end stack Achieving state-of-the-art performance for its size across math, code and reasoning Built using the same tools we put in your hands, from environments & evals, RL frameworks, sandboxes & more

Prime Intellect

1,150,419 views • 9 months ago

Lecture 10 of my course! Nominally on regularization in RL, so I discuss the evolving role of the KL penalty in RL, but also a set of nice RL papers that explain what RL helps models generalize better than SFT -- with theory supporting it. When going through these, it's so interesting how seasonal problems in ML are. Lots of problems from controlling reward models overopt will rhyme as we try to control rubrics for agents. 00:00 Intro & the role of regularization 02:50 The KL penalty in RL 10:53 RL as a reverse KL loss 21:09 Why RL generalizes better than SFT 25:15 Other regularization tools Just a few videos left as I get to the end of the course. Thanks all, and keep sending questions. Spread the word if you have a second.

Lecture 10 of my course! Nominally on regularization in RL, so I discuss the evolving role of the KL penalty in RL, but also a set of nice RL papers that explain what RL helps models generalize better than SFT -- with theory supporting it. When going through these, it's so interesting how seasonal problems in ML are. Lots of problems from controlling reward models overopt will rhyme as we try to control rubrics for agents. 00:00 Intro & the role of regularization 02:50 The KL penalty in RL 10:53 RL as a reverse KL loss 21:09 Why RL generalizes better than SFT 25:15 Other regularization tools Just a few videos left as I get to the end of the course. Thanks all, and keep sending questions. Spread the word if you have a second.

Nathan Lambert

89,468 views • 1 month ago

This is a 20-minute unedited on/off-gantry run from today. What's special about it is that no gains were tuned on the robot before we started running the policy. This is the DVBF-based motion policy we've been working on, which is turning out to be quite special. The goal of using DVBF is to ensure policies can be deployed on many robots at once without manual gain adjustment before the policy runs reliably on each robot.

This is a 20-minute unedited on/off-gantry run from today. What's special about it is that no gains were tuned on the robot before we started running the policy. This is the DVBF-based motion policy we've been working on, which is turning out to be quite special. The goal of using DVBF is to ensure policies can be deployed on many robots at once without manual gain adjustment before the policy runs reliably on each robot.

Sankaet

36,570 views • 10 months ago

Before humanoid robots are ready for homes, they need to pass the "Home Alone test," according to Standard Bots CEO Evan Beard. "We think maybe humanoids are right for the home. But then you have the safety issue. When it's going up the stairs, you need two five-year-olds to be able to jump on the back of it, pull it, or try to trip it. It's the Home Alone test." "You see RL robots and it looks so impressive. I think the average person thinks 'Oh yeah, robotics is solved.' But to beat the Home Alone test, RL is not quite there and it's hard to simulate all the things that an eight-year-old could do to it. So you can't even do the RL right now. It's an unsolved problem... I think it's going to be years away from being solved in the home."

Before humanoid robots are ready for homes, they need to pass the "Home Alone test," according to Standard Bots CEO Evan Beard. "We think maybe humanoids are right for the home. But then you have the safety issue. When it's going up the stairs, you need two five-year-olds to be able to jump on the back of it, pull it, or try to trip it. It's the Home Alone test." "You see RL robots and it looks so impressive. I think the average person thinks 'Oh yeah, robotics is solved.' But to beat the Home Alone test, RL is not quite there and it's hard to simulate all the things that an eight-year-old could do to it. So you can't even do the RL right now. It's an unsolved problem... I think it's going to be years away from being solved in the home."

TBPN

32,366 views • 3 months ago

I always dreamed of AGI as a wise advisor for humanity. Although LLMs are great for coding & knowledge work, I wouldn’t trust them to give me advice on my career, business strategy, or policy preferences. How can we build AI systems optimized for wisdom? At Mantic we believe the unlock is prediction: predicting world events as accurately as possible, and hill-climbing this single metric. Today we share some recent progress on the Thinking Machines website, having found Tinker a great platform for our RL experiments. TL;DR: We RL-tune gpt-oss-120b to become a better forecaster than any other model. Having good scaffolding is a prerequisite. A fun result: our tuned model + Grok are decorrelated from the other best models, and so are the most indispensable when picking a team.

I always dreamed of AGI as a wise advisor for humanity. Although LLMs are great for coding & knowledge work, I wouldn’t trust them to give me advice on my career, business strategy, or policy preferences. How can we build AI systems optimized for wisdom? At Mantic we believe the unlock is prediction: predicting world events as accurately as possible, and hill-climbing this single metric. Today we share some recent progress on the Thinking Machines website, having found Tinker a great platform for our RL experiments. TL;DR: We RL-tune gpt-oss-120b to become a better forecaster than any other model. Having good scaffolding is a prerequisite. A fun result: our tuned model + Grok are decorrelated from the other best models, and so are the most indispensable when picking a team.

Toby Shevlane

183,370 views • 6 months ago