Skip to content

Dataset · harness-hub

traj2envs Reinforcement Learning Tasks

harness-hub/traj2envs-rl

A set of reinforcement learning tasks based on trajectories.

What's inside

Contains 270 Harbor tasks in parquet format with metadata, environment docker images, and solution success rates. Real interaction trajectories served as the source. The dataset volume is less than 1 thousand items.

What you can do with it

  • Training agents using Reinforcement Learning methods
  • Evaluating and benchmarking models for operating in isolated environments
Size
up to 1K rows
Downloads this month
16K
Likes
3

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("harness-hub/traj2envs-rl")