Semantic spatial verification reduces false object claims in multimodal AI
Semantic-Spatial Agreement Verification for Mitigating Object Hallucination in Multimodal Large Language Models
Computer Vision and Pattern Recognition
Summary
Multimodal AI models that describe images sometimes mention things that aren't actually there, which can cause problems in sensitive areas like medicine or safety. The authors propose a new method called Semantic-Spatial Agreement Verification (SSAV) that checks if an object really appears in an image by seeing if different ways of asking about it point to the same place visually. This approach does not require retraining the model and helps reduce mistakes where the AI hallucinates objects. Tests show SSAV improves accuracy in identifying real objects and lowers errors across several image question-answering benchmarks.
What this means in practice
- •For assistive technology developers: Reduce incorrect object mentions in AI-generated descriptions for users with visual impairments to improve trust and safety.
- •For automated medical support teams: Verify visual claims in AI-assisted medication or monitoring systems to avoid dangerous hallucination errors.
Authors
Ziheng Ren, Qian Gao, Jun Fan, Guohui Ding, Zhenyu Yang, Yuteng Xiao
Abstract
Multimodal large language models generate natural-language responses from visual inputs, yet may mention objects absent from an image. In medication assistance, accessible perception, and environmental decision-making, such hallucinations can create real-world safety risks. We propose Semantic-Spatial Agreement Verification (SSAV), a training-free method for verifying object claims. A visually grounded claim should remain stable across semantically equivalent queries and repeatedly localize to the same image region. SSAV aggregates multiple prompts to estimate semantic support and reduce sensitivity to query wording. Query-Induced Regional Verification (QIRV) combines cross-query region persistence, spatial overlap, and relative candidate dominance to identify isolated high responses and dispersed localizations. A geometric mean fuses semantic and spatial evidence, lowering the verification score when either branch lacks support. Experiments on three base models and multiple evaluation protocols show that SSAV effectively mitigates object hallucination. On LLaVA-1.5-7B, accuracy averaged across COCO, A-OKVQA, and GQA improves by 1.81 and 3.17 percentage points under POPE Popular and Adversarial, respectively, while CHAIRs decreases from 49.40% to 32.80%. These results show that cross-query semantic stability and regional consistency provide interpretable external visual evidence for object claims.