Preprint
Aug 2026
Learning from Failures: Retrieval-Centric CoT via Hard Negatives for Unified Multimodal Retrieval
This work introduces UniME-R1, an embedder-adviser framework that learns to reason over initially retrieved candidates and generate Retrieval-Centric Chain-of-Thought (RC-CoT), and aligns the adviser with retrieval outcomes through supervised learning and retrieval-oriented reinforcement learning.
Ze-Long Sun, Jun Wang, Kaicheng Yang et al.
· 0 citations