OzLabs/

gatling-tier1-338k

	Gatling Tier I: Standard Utility Dataset (338K Samples)




	
	Dataset Summary

Gatling Tier I is a large-scale dataset of benign tool-use execution plans for training energy-based security models for agentic systems. This dataset represents the "safe valley" of standard utility—typical, legitimate patterns of AI agent tool interactions that should receive low energy scores in Gatling's Joint-Embedding Predictive Architecture (JEPA). This dataset is part of the Gatling-10M… See the full description on the dataset page: https://huggingface.co/datasets/OzLabs/gatling-tier1-338k.