Prompt Sensitivity in Vision-Language Grounding: How Small Changes in Wording Affect Object Detection

arXiv:2604.17126v1 Announce Type: new Abstract: Vision-language models enable open-vocabulary object grounding through natural language queries, under the implicit assumption that semantically equivalent descriptions yield consistent outputs. We examine this assumption using a controlled pipeline combining DETR for object proposals with CLIP for language-conditioned selection on 263 COCO val2017 images. We find that overlapping prompts such as "a person," "a human," and "a pedestrian" frequently select different instances, with mean instability of 2.11 distinct selections across six prompts. PCA analysis shows this variability is structured and directional, not random. Prompt ensembling does not improve quality and often shifts selections toward generic regions. We further show that text embedding proximity explains only 34% of grounding disagreement (r = -0.58), confirming that instability arises from the argmax selection mechanism rather than text-level distances alone.

Related

• HaloProbe: Bayesian Detection and Mitigation of Object Hallucinations in Vision-Language Models
• Cross-Attentive Multiview Fusion of Vision-Language Embeddings
• Through the Magnifying Glass: Adaptive Perception Magnification for Hallucination-Free VLM Decoding
• Latent Anomaly Knowledge Excavation: Unveiling Sparse Sensitive Neurons in Vision-Language Models
• HTDC: Hesitation-Triggered Differential Calibration for Mitigating Hallucination in Large Vision-Language Models
• Tarot-SAM3: Training-free SAM3 for Any Referring Expression Segmentation

Source: arXiv cs.CV | 2026-04-21