Video wird geladen...

Video konnte nicht geladen werden

Beim Laden dieses Videos ist ein Problem aufgetreten. Dies könnte an einem vorübergehenden Netzwerkproblem liegen oder das Video ist möglicherweise nicht verfügbar.

Turned Scott Wu’s pub card game into a Math RL environment Simple rules and rewards, and plenty of headroom for small models to hillclimb on and built on OpenEnv 👀

Adithya S K

16,696 subscribers

17,833 Aufrufe • vor 15 Tagen •via X (Twitter)

Wissenschaft & Technologie Gaming Bildung

Anya Rossi• Live Now

Private livecam show

16 Kommentare

Profilbild von Adithya S K

Adithya S Kvor 15 Tagen

> 4,000 train + 200 test deals, unlimited fresh ones, all solver-checked ps: evals subjected to inference speed Play it or train on it:

Profilbild von sankalp

sankalpvor 15 Tagen

😭

Profilbild von TaraT

TaraTvor 15 Tagen

lmao

Profilbild von Umar Ghani

Umar Ghanivor 15 Tagen

Cool stuff

Profilbild von QM DOMAINS | مال.com

QM DOMAINS | مال.comvor 15 Tagen

cool

Profilbild von Sabhya Chhabria

Sabhya Chhabriavor 15 Tagen

Super cool!

Profilbild von Xiangyi Li

Xiangyi Livor 15 Tagen

lol man you are a genius

Profilbild von manv

manvvor 14 Tagen

😭

Profilbild von Sergio Paniego

Sergio Paniegovor 15 Tagen

wtf 💀haha

Profilbild von xtamales

xtamalesvor 15 Tagen

cool format!

Profilbild von Alex Xi

Alex Xivor 14 Tagen

what were the results

Profilbild von Florian Findeis

Florian Findeisvor 14 Tagen

Do models only get a reward for hitting the number exactly, or also for getting close?

Profilbild von Alex

Alexvor 14 Tagen

Sad reality, any verifiable human skill will get turned into a RL environment and get hillclimbed to oblivion

Profilbild von Aayushman Singh

Aayushman Singhvor 14 Tagen

Hold out card combinations to see if the gains survive new hands.

Profilbild von Solomon

Solomonvor 15 Tagen

Solver-checked synthetic environments with fast step times are the holy grail for RL on smaller weights. Are you using sparse terminal rewards or shaping intermediate trick margins to prevent policy collapse during early exploration?

Profilbild von Aadithya Sathish Kumar

Aadithya Sathish Kumarvor 15 Tagen

@matiks_play

Ähnliche Videos

Introducing RL Environment Explorer 🧭 There are now 12M+ RL task entries across RL environments on the Hugging Face Hub. You can now explore, visualize, and run RL environments across: OpenEnv • Harbor • MiMo • Verifiers • NeMo Gym All in one place

Introducing RL Environment Explorer 🧭 There are now 12M+ RL task entries across RL environments on the Hugging Face Hub. You can now explore, visualize, and run RL environments across: OpenEnv • Harbor • MiMo • Verifiers • NeMo Gym All in one place

Adithya S K

27,777 Aufrufe • vor 1 Monat

To solve AGI, we must first solve Geoguessr For that I built vlm-gym, a simple RL gym written in scratch, in JAX for Qwen3VL-4B (released yesterday) And added Geospot, a RL environment for geolocation and learned VLMs can learn how to geoguess. More:

To solve AGI, we must first solve Geoguessr For that I built vlm-gym, a simple RL gym written in scratch, in JAX for Qwen3VL-4B (released yesterday) And added Geospot, a RL environment for geolocation and learned VLMs can learn how to geoguess. More:

Surya Dantuluri

140,580 Aufrufe • vor 1 Jahr

Advice for AI engineers 💡 Browser control is possible with a Small Model, like LFM2-350M by Liquid AI . Here's a 60-minute deep dive on how to fine-tune with RL and OpenEnv by Hugging Face Enjoy ↓

Advice for AI engineers 💡 Browser control is possible with a Small Model, like LFM2-350M by Liquid AI . Here's a 60-minute deep dive on how to fine-tune with RL and OpenEnv by Hugging Face Enjoy ↓

Pau Labarta Bajo

16,169 Aufrufe • vor 6 Monaten

hillclimb (hillclimb) is the human superintelligence community, dedicated to building golden datasets for AGI. Starting with math, their team of IMO medalists, lean experts, PhDs is designing RL environments for Nous Research.

hillclimb (hillclimb) is the human superintelligence community, dedicated to building golden datasets for AGI. Starting with math, their team of IMO medalists, lean experts, PhDs is designing RL environments for Nous Research.

Y Combinator

912,747 Aufrufe • vor 11 Monaten

A simple game of hide and seek turned into something no one was prepared for 😂

A simple game of hide and seek turned into something no one was prepared for 😂

Styke

1,564,748 Aufrufe • vor 5 Monaten

Releasing SmolDataEnvs 🤗 5K+ Verifiable RL Environment tasks for hill-climbing small models in code and data science. Completely open source: environments, evals, training

Releasing SmolDataEnvs 🤗 5K+ Verifiable RL Environment tasks for hill-climbing small models in code and data science. Completely open source: environments, evals, training

Adithya S K

55,579 Aufrufe • vor 1 Monat

Points are a thing of the past. Welcome to the new age of credit card rewards. We’re spotlighting bitcoin as the new standard for credit card rewards with a campaign we built with 150+ billboards across NYC and Austin. Unlike points, bitcoin doesn’t expire, it doesn’t get devalued by fine print, and it’s built to grow with you. Cardholders earning bitcoin for over 1 year have seen their rewards appreciate 277% on average. The Gemini Credit Card turns every purchase into bitcoin with no annual fee. 🔗 Go further with bitcoin rewards:

Points are a thing of the past. Welcome to the new age of credit card rewards. We’re spotlighting bitcoin as the new standard for credit card rewards with a campaign we built with 150+ billboards across NYC and Austin. Unlike points, bitcoin doesn’t expire, it doesn’t get devalued by fine print, and it’s built to grow with you. Cardholders earning bitcoin for over 1 year have seen their rewards appreciate 277% on average. The Gemini Credit Card turns every purchase into bitcoin with no annual fee. 🔗 Go further with bitcoin rewards:

Gemini

42,553 Aufrufe • vor 1 Jahr

We (nico and I) turned Warcraft 3 Frozen Throne into a real RL env wc3env is a fast, concurrent, and deterministic environment for training and evaluating models/agents Frontier agents do ok in isolated battles but fall apart in full 1v1s due to limitations in spatial reasoning, intelligence, and speed Open source and writeup ↓

We (nico and I) turned Warcraft 3 Frozen Throne into a real RL env wc3env is a fast, concurrent, and deterministic environment for training and evaluating models/agents Frontier agents do ok in isolated battles but fall apart in full 1v1s due to limitations in spatial reasoning, intelligence, and speed Open source and writeup ↓

Peter Wang

48,447 Aufrufe • vor 1 Monat

This is a death game built on math. Get the digits wrong, and everyone’s dies.

This is a death game built on math. Get the digits wrong, and everyone’s dies.

🧌JungleRock

12,970 Aufrufe • vor 1 Monat

Introducing arkenOS — an autonomous training gym that takes any RL agent from a goal to a proven, deployable policy. We are automating complex RL research, enabling high-velocity training and rigorous, diverse environment design and testing right on your local hardware. Out of the box. Our bet is simple: The future of AI will be built by companies giving millions of agents millions of simulated environments to adapt and learn cause and effect in, and survive. This continuous, autonomous selection will create a digital Cambrian explosion, distilling evolutionary trial and error into a foundational specialized intuition.

Introducing arkenOS — an autonomous training gym that takes any RL agent from a goal to a proven, deployable policy. We are automating complex RL research, enabling high-velocity training and rigorous, diverse environment design and testing right on your local hardware. Out of the box. Our bet is simple: The future of AI will be built by companies giving millions of agents millions of simulated environments to adapt and learn cause and effect in, and survive. This continuous, autonomous selection will create a digital Cambrian explosion, distilling evolutionary trial and error into a foundational specialized intuition.

Dev Shah

18,099 Aufrufe • vor 3 Monaten

Our Inference API is live. Use it to evaluate environments on the Hub. + Simple `prime env eval` cmd + One endpoint for all models + 56 models and growing + Unified billing with compute + Viewer for rewards and rollouts + Share results on Hub

Our Inference API is live. Use it to evaluate environments on the Hub. + Simple `prime env eval` cmd + One endpoint for all models + 56 models and growing + Unified billing with compute + Viewer for rewards and rollouts + Share results on Hub

Prime Intellect

53,216 Aufrufe • vor 1 Jahr

This mom turned a simple chair into a roller coaster just to make her daughter smile. ❤️ A reminder that childhood isn't built on expensive toys. It's built on love, creativity, and the people who make ordinary moments feel magical.

This mom turned a simple chair into a roller coaster just to make her daughter smile. ❤️ A reminder that childhood isn't built on expensive toys. It's built on love, creativity, and the people who make ordinary moments feel magical.

Ersin

67,151 Aufrufe • vor 5 Monaten

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

1,155,014 Aufrufe • vor 11 Monaten

Clavicular checked a guy for trying to hit on his girl and it almost turned into a fight👀

Clavicular checked a guy for trying to hit on his girl and it almost turned into a fight👀

⟡

26,744 Aufrufe • vor 8 Monaten

A conversation on the optimal reward for coding agents, infinite context models, and real-time RL

A conversation on the optimal reward for coding agents, infinite context models, and real-time RL

Cursor

320,339 Aufrufe • vor 1 Jahr

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 Aufrufe • vor 7 Monaten

OpenEnv has a new home: starting today, it's coordinated by a committee that includes Meta-PyTorch, Reflection, Unsloth, Modal, Prime Intellect, Nvidia, Mercor, Fleet AI, and Hugging Face frontier labs train their models and their harnesses together. Claude knows Claude Code. GPT-5.5 knows Codex. that's not an accident, it's training. open-source models deserve the same magic, but pulling that off requires infrastructure that belongs to everyone, not one lab OpenEnv is that layer. one api, any harness, any trainer, any environment rewards and training loops stay in TRL, Unsloth, wherever you already work. OpenEnv is the socket they all plug into get involved!

OpenEnv has a new home: starting today, it's coordinated by a committee that includes Meta-PyTorch, Reflection, Unsloth, Modal, Prime Intellect, Nvidia, Mercor, Fleet AI, and Hugging Face frontier labs train their models and their harnesses together. Claude knows Claude Code. GPT-5.5 knows Codex. that's not an accident, it's training. open-source models deserve the same magic, but pulling that off requires infrastructure that belongs to everyone, not one lab OpenEnv is that layer. one api, any harness, any trainer, any environment rewards and training loops stay in TRL, Unsloth, wherever you already work. OpenEnv is the socket they all plug into get involved!

Sergio Paniego

32,760 Aufrufe • vor 5 Monaten

Clavicular checked a guy for hitting on his girl and it almost turned into a fight 👀

Clavicular checked a guy for hitting on his girl and it almost turned into a fight 👀

KickChamp👑

195,833 Aufrufe • vor 1 Jahr

Went to check on my boys earlier today, Everywhere good. Over 10 years in the game for Rams and a few cows but now into the Cow game fully. My rams should arrive in January in preparation for Ileya. Plenty investment, Plenty Risk, Plenty Patience and Plenty Profit, If Your mind can carry all the “plenty” together. 😁

Went to check on my boys earlier today, Everywhere good. Over 10 years in the game for Rams and a few cows but now into the Cow game fully. My rams should arrive in January in preparation for Ileya. Plenty investment, Plenty Risk, Plenty Patience and Plenty Profit, If Your mind can carry all the “plenty” together. 😁

Jubril A. Gawat

125,516 Aufrufe • vor 11 Monaten

trained my own with trl + openenv blog with open artifacts (code, models, dataset, rl env...) soon!

trained my own with trl + openenv blog with open artifacts (code, models, dataset, rl env...) soon!

Sergio Paniego

90,257 Aufrufe • vor 2 Monaten