Skip to content

Dataset · nvidia

RL Dataset for AI Agents in Terminal

nvidia/Nemotron-RL-Agentic-Terminal-Pivot-v1

English command-line action trajectories from NVIDIA for reinforcement learning of LLM agents.

What's inside

This NVIDIA dataset contains 10k to 100k English training examples for RL training of command-line agents in the terminus_judge environment of the NeMo Gym framework. Each record represents a decision point from a successful trajectory: task instructions and terminal interaction history, including previous agent actions and console output.

What you can do with it

  • RL fine-tuning of LLM agents for autonomous command-line tasks
  • Training planning and decision-making models in interactive environments using NeMo Gym
Tasks
Text generation
Languages
English
Size
10K–100K rows
Downloads this month
2.3K
Likes
33

How to load it

from datasets import load_dataset

ds = load_dataset("nvidia/Nemotron-RL-Agentic-Terminal-Pivot-v1")