Search papers, labs, and topics across Lattice.
This paper identifies and addresses a critical failure mode in vision-language models (VLMs) known as Contextual Variable Overestimation (CVE), where models prioritize dominant visual-textual cues over sparse yet crucial contextual variables in cross-market preference prediction. By developing GeoReward, a novel reward model that incorporates Market-Aware Retrieval Augmentation, Context-Guided Visual Modulation, and Selective Sensitivity Loss, the authors demonstrate improved performance in predicting advertisement image preferences across diverse geographic markets. Experiments show that GeoReward effectively mitigates CVE, outperforming existing baselines and providing a robust solution for applications sensitive to regional variations in decision-making.
VLMs can overlook critical regional cues, but GeoReward effectively recalibrates their focus, leading to superior cross-market preference predictions.
Vision-language models excel in many multimodal tasks but remain prone to a subtle yet impactful failure mode: they tend to overestimate dominant visual-textual cues while underestimating sparse but decision-critical contextual variables. This issue, which we term Contextual Variable Overestimation (CVE), becomes particularly evident in real-world applications such as predicting advertisement image preferences across diverse geographic markets. For instance, when a VLM is asked to choose between two product images tailored for different countries, it often defaults to a consistent output, ignoring ground-truth regional variations. This collapse occurs because pervasive high-volume signals, such as product attributes and dense image patches, overwhelm the few but critical tokens that encode market-specific context. To address CVE, we first collect a new multimodal dataset of real advertising creatives and their click-through performance across multiple countries. We then introduce GeoReward, a reward model designed to predict ad image preferences across diverse geographic markets. GeoReward integrates three purpose-built mechanisms: (1) Market-Aware Retrieval Augmentation, (2) Context-Guided Visual Modulation, (3) Selective Sensitivity Loss. Furthermore, we demonstrate how GeoReward can guide the fine-tuning of RL for a VLM to generate background designs for text-to-image models, producing market-aware advertising creatives. Experiments validate that our framework mitigates CVE and outperforms existing baselines. This work not only diagnoses a systematic bias in VLMs toward dominant perceptual features but also delivers a targeted solution for applications where sparse contextual variables govern decision-making.
