alibayram/

gemini-3.1-pro-hard-high-reasoning

	Dataset Card for Gemini-3.1-Pro-Ultra-Reasoning-5.6M




	
	Dataset Details




	
	Dataset Description

This dataset represents the frontier of synthetic reasoning data, generated by Gemini 3.1 Pro (High Reasoning variant). While smaller in total token volume than its predecessors (5.6M tokens), this corpus prioritizes logical density and multi-step verification. The move to the 3.1 architecture provides a measurable leap in "System 2" thinking. Unlike standard models… See the full description on the dataset page: https://huggingface.co/datasets/alibayram/gemini-3.1-pro-hard-high-reasoning.