On Data Engineering for Scaling LLM Terminal Capabilities

TL;DR

This paper systematically studies data engineering practices for large language model terminal capabilities, introduces a synthetic task pipeline, and trains models that significantly improve terminal task performance, opening resources for further research.

Contribution

It presents a novel synthetic task generation pipeline and a comprehensive analysis of data strategies, leading to improved terminal models and an open-source dataset.

Findings

Models trained on the new dataset outperform previous versions.

Scaling model size improves terminal task accuracy.

Open-sourcing datasets and checkpoints facilitates future research.

Abstract

Despite rapid recent progress in the terminal capabilities of large language models, the training data strategies behind state-of-the-art terminal agents remain largely undisclosed. We address this gap through a systematic study of data engineering practices for terminal agents, making two key contributions: (1) Terminal-Task-Gen, a lightweight synthetic task generation pipeline that supports seed-based and skill-based task construction, and (2) a comprehensive analysis of data and training strategies, including filtering, curriculum learning, long context training, and scaling behavior. Our pipeline yields Terminal-Corpus, a large-scale open-source dataset for terminal tasks. Using this dataset, we train Nemotron-Terminal, a family of models initialized from Qwen3(8B, 14B, 32B) that achieve substantial gains on Terminal-Bench 2.0: Nemotron-Terminal-8B improves from 2.5% to 13.0% Nemotron-Terminal-14B improves from 4.0% to 20.2%, and Nemotron-Terminal-32B improves from 3.4% to 27.4%, matching the performance of significantly larger models. To accelerate research in this domain, we open-source our model checkpoints and most of our synthetic datasets at https://huggingface.co/collections/nvidia/nemotron-terminal.