Multimodal contrastive models are widely used for bidirectional image-text retrieval, but their rankings remain vulnerable to attraction bias, where certain candidates become hubs for many unrelated queries. Existing training-free normalization methods suppress such hubs, but may also penalize candidates that are genuinely relevant. We propose GLOBE, a training-free plug-in module that combines global-local bias calibration with graph-based semantic evidence to suppress unsupported hubs while preserving true matches. Across MS-COCO and Flickr30K, GLOBE improves diverse frozen retrieval backbones, achieving up to 7.9% absolute Recall@1 gain over the original model and up to 5.2% over strong training-free baselines. We further construct IRB-COCO-500, a mixed real-and-synthetic benchmark with AI-generated near-positive distractors, and show that GLOBE improves retrieval robustness in this challenging setting.

Reward Models (RMs) are critical components in the Reinforcement Learning from Human Feedback (RLHF) pipeline, directly determining the alignment quality of Large Language Models (LLMs). Recently, Generative Reward Models (GRMs) have emerged as a superior paradigm, offering higher interpretability and stronger generalization than traditional scalar RMs. However, existing methods for GRMs focus primarily on outcome-level supervision, neglecting analytical process quality, which constrains their potential. To address this, we propose ReflectRM, a novel GRM that leverages self-reflection to assess analytical quality and enhance preference modeling. ReflectRM is trained under a unified generative framework for joint modeling of response preference and analysis preference. During inference, we use its self-reflection capability to identify the most reliable analysis, from which the final preference prediction is derived. Experiments across four benchmarks show that ReflectRM consistently improves performance, achieving an average accuracy gain of +3.7 on Qwen3-4B. Further experiments confirm that response preference and analysis preference are mutually reinforcing. Notably, ReflectRM substantially mitigates positional bias, yielding +10.2 improvement compared with leading GRMs and establishing itself as a more stable evaluator. Our code is available at https://github.com/yuliangCarmelo/ReflectRM.