Researcher, ByteDance Inc.
1 paper at NeurIPS 2025
We propose Q-Insight, a reinforcement learning-based IQA approach using limited annotations, achieving superior performance and enhanced visual reasoning capabilities.