How Many Visual Levers Drive Urban Perception? Interventional Counterfactuals via Multiple Localised Edits

arXiv:2604.22103v1 Announce Type: cross Abstract: Street-view perception models predict subjective attributes such as safety at scale, but remain correlational: they do not identify which localized visual changes would plausibly shift human judgement for a specific scene. We propose a lever-based interventional counterfactual framework that recasts scene-level explainability as a bounded search over structured counterfactual edits. Each lever specifies a semantic concept, spatial support, intervention direction, and constrained edit template. Candidate edits are generated through prompt-conditioned image editing and retained only if they satisfy validity checks for same-place preservation, locality, realism, and plausibility. In a pilot across 50 scenes from five cities, the framework reveals preliminary proxy-based directional patterns and a practical failure taxonomy under prompt-only editing, with Mobility Infrastructure and Physical Maintenance showing the largest auxiliary safety shifts. Human pairwise judgements remain the ground-truth endpoint for future validation.

Related

• Bias Redistribution in Visual Machine Unlearning: Does Forgetting One Group Harm Another?
• Visual Adversarial Attack on Vision-Language Models for Autonomous Driving
• Cross-Modal Attention Analysis and Optimization in Vision-Language Models: A Study on Visual Reliability
• Faithful GRPO: Improving Visual Spatial Reasoning in Multimodal Language Models via Constrained Policy Optimization

Source: arXiv cs.CV | 2026-04-27