Adithya S K

@adithya_s_k

Scaling RL Envs @huggingface 🤗 • Founded @cognitivelab_ai Prev : Research @MSFTResearch • ML @apple • 22

following
followers

Technology

Two OpenEnv environments open on Hugging Face with 12M-plus tasks

A user says OCR spans 1M-plus document pages in 22 languages and speech recognition spans 102 languages.

2 Sources, 6d ago, first seen 6d ago

Two open-source reinforcement-learning environments offer 12 million-plus OCR and speech-recognition tasks

A post introducing the environments says both are available on Hugging Face, with speech recognition covering 102 languages.

2 Sources, 6d ago, first seen 6d ago

Technology

RL Environment Explorer introduced for exploring and running RL environments

The introductory post says the Hugging Face Hub has 12 million-plus RL task entries across RL environments.

2 Sources, 13d ago, first seen 13d ago

RL Environment Explorer puts reinforcement-learning environments in one place

A post announcing the explorer says the Hugging Face Hub has 12M-plus reinforcement-learning task entries.

5 Sources, 12d ago, updated 12d ago

OpenAI releases results from its internal AI math model on GitHub

OpenAI says it drew on an independent mathematics advisory group’s advice about how to release the results.

435 Sources, 9d ago, updated 9d ago

Hugging Face adds shared discovery for reinforcement-learning environments

Hugging Face’s framework tags identify compatible tasksets, while execution stays with the framework and its runtime.

5 Sources, 15d ago, first seen 15d ago

Technology

RL environments reportedly now have a home on the Hugging Face Hub

In a post announcing the change, Ben Burtenshaw said reinforcement learning environments should work more like datasets on the Hub, which he said already stores, versions, gates, and previews data.

3 Sources, 15d ago, first seen 15d ago

Capture proxy enables reinforcement learning for open models across coding-agent harnesses

A post says the OpenEnv proxy records token IDs and log probabilities while harness runs remain unmodified.

22 Sources, 14d ago, updated 14d ago

Technology

OpenEnv gets a capture proxy for training open models across coding harnesses

A post says the proxy records each model call’s exact token IDs and log probabilities while the harnesses run unmodified.

7 Sources, 18d ago, updated 18d ago

Technology

TaskSmith released to generate reinforcement-learning environments from code

The release post describes TaskSmith as a specialized orchestrator, rather than a general coding agent with a long prompt.

2 Sources, 20d ago, first seen 20d ago

TaskSmith released to generate reinforcement-learning environments from code

Its developer describes TaskSmith as a specialized orchestrator, not a general coding agent with a long prompt. Each stage is designed to turn a pull request into a reinforcement-learning environment.

2 Sources, 20d ago, first seen 20d ago

989 environments reportedly used to train a distilled 9B model

A user who examined the release says reward scoring isn't self-contained: general environments need a judge set up, while web-development environments rely on a grader service and a vision-language model.

17 Sources, 20d ago, updated 20d ago

SmolDataEnvs launches with 5,000-plus reinforcement-learning tasks

The release announcement describes verifiable tasks for improving small models in code and data science, with environments, evaluations and training all open source.

7 Sources, 24d ago, updated 24d ago

Hugging Face training loop reportedly reaches the same reward in 53 minutes, down from 3 hours 27

A user describes a reinforcement learning setup that keeps training, model serving and storage on Hugging Face, syncing only small adapter updates through Buckets.

2 Sources, 32d ago, first seen 32d ago

Brendan Foody Recommends Edward's LoRA Post-Training Blog

Mercor CEO highlights Edward's blog for open-source model post-training.

3 Sources, 46d ago, first seen 46d ago

Hugging Face Engineer Releases Open Write-Up on Code-Based Painting Training

Post details environment, hand-rated pool, runs, and all generated paintings from the project.

4 Sources, 47d ago, first seen 47d ago

Researcher Announces Upcoming GeoGuesser RL Open Source Release

Tweet from Adithya S K at Hugging Face promises full reproduction materials for training a VLM on the game.

2 Sources, 48d ago, first seen 48d ago

Featured in

Two OpenEnv environments open on Hugging Face with 12M-plus tasks

Two open-source reinforcement-learning environments offer 12 million-plus OCR and speech-recognition tasks

RL Environment Explorer introduced for exploring and running RL environments

RL Environment Explorer puts reinforcement-learning environments in one place

OpenAI releases results from its internal AI math model on GitHub

Hugging Face adds shared discovery for reinforcement-learning environments

RL environments reportedly now have a home on the Hugging Face Hub

Capture proxy enables reinforcement learning for open models across coding-agent harnesses

OpenEnv gets a capture proxy for training open models across coding harnesses

TaskSmith released to generate reinforcement-learning environments from code

TaskSmith released to generate reinforcement-learning environments from code

989 environments reportedly used to train a distilled 9B model

SmolDataEnvs launches with 5,000-plus reinforcement-learning tasks

Hugging Face training loop reportedly reaches the same reward in 53 minutes, down from 3 hours 27

Brendan Foody Recommends Edward's LoRA Post-Training Blog

Hugging Face Engineer Releases Open Write-Up on Code-Based Painting Training

Researcher Announces Upcoming GeoGuesser RL Open Source Release

Featured in

Two OpenEnv environments open on Hugging Face with 12M-plus tasks

Two open-source reinforcement-learning environments offer 12 million-plus OCR and speech-recognition tasks

RL Environment Explorer introduced for exploring and running RL environments

RL Environment Explorer puts reinforcement-learning environments in one place

OpenAI releases results from its internal AI math model on GitHub

Hugging Face adds shared discovery for reinforcement-learning environments

RL environments reportedly now have a home on the Hugging Face Hub

Capture proxy enables reinforcement learning for open models across coding-agent harnesses

OpenEnv gets a capture proxy for training open models across coding harnesses

TaskSmith released to generate reinforcement-learning environments from code

TaskSmith released to generate reinforcement-learning environments from code

989 environments reportedly used to train a distilled 9B model

SmolDataEnvs launches with 5,000-plus reinforcement-learning tasks

Hugging Face training loop reportedly reaches the same reward in 53 minutes, down from 3 hours 27

Brendan Foody Recommends Edward's LoRA Post-Training Blog

Hugging Face Engineer Releases Open Write-Up on Code-Based Painting Training

Researcher Announces Upcoming GeoGuesser RL Open Source Release