This repository trains LLMs to perform multi-turn Tool-Integrated Reasoning (TIR) with RL, where LLMs iteratively generate code, execute it, and think upon the execution results. This capability ...
Classification (TF-Cls) 'Clear', 'Closed', 'Broken', 'Blur' 6,247 3632 × 2760 4,687:561:999(75%:9%:16%) Object Detection (TF-Det) Inside, Middle, Outside Rings 4,736 ...