xiuhuywh/DRIM-VisualReasonHard

This repository contains the RL training datasets used in the paper Deep But Reliable: Advancing Multi-turn Reasoning for Thinking with Images