Hugging Face

450+ multi-step data-analysis tasks combining structured records with unstructured docs, requiring an agent to plan, query, and reason in stages.

🍷 FineWeb

JAT Dataset

📚 FineWeb-Edu

Dataset Card for HuggingFaceFW/finephrase

Dataset Card for "arxiv-metadata-oai-snapshot"

OpenEnv port of AI4Finance's FinRL - financial-market simulators (NASDAQ-100, DJIA, S&P 500, HSI, CSI 300) for training trading agents with DRL.

Stateful Jupyter-notebook environment backed by E2B sandboxes; the model edits and executes cells turn-by-turn, building up a notebook that solves a data-science task.

HuggingFace LeRobot's EnvHub - a one-line `from_hub()` loader for sim and real-robot environments, hosted on the Hub as standard repos, compatible with the LeRobot gym/dataset/policy stack.

OpenEnv wrapper around the Arcade Learning Environment - the classic 57-game Atari 2600 benchmark suite that defined deep RL.

OpenEnv port of DeepMind's OpenSpiel - 70+ classic games (Chess, Go, Poker, Hanabi, Liar's Dice, etc.) with single-agent and multi-agent RL APIs.

Single-tool (tmux session) terminal coding environment in the OpenEnv standard, backed by Hugging Face Sandbox containers - the OpenEnv port of the Terminus-2 design.

OpenEnv port of Lucas Alegre's SUMO-RL - multi-intersection traffic-signal-control environments driven by the SUMO microscopic traffic simulator.

OpenEnv port of ServiceNow's BrowserGym - a Playwright-backed browser environment exposing WebArena, MiniWoB, WorkArena, etc. through the standard OpenEnv API.

OpenEnv port of DeepMind's dm_control - MuJoCo-powered continuous control tasks (cartpole, walker, humanoid, dog, quadruped, manipulation) for embodied RL.

OpenEnv port of TextArena - a collection of competitive text games (Wordle, Snake, Tic-Tac-Toe, etc.) for evaluating LLM reasoning under interactive rules.

HuggingFace Jupyter-Agent - an LLM scaffold that drives a real E2B-backed Jupyter notebook for data-analysis tasks, with execution, plots, and tool-calling baked in.