Now You See the Hate: Adaptive View Retrieval for Hidden Hateful Illusions

2026-07-21

AI Summary

Existing vision-language multimodal safety systems struggle to detect hate content concealed within visual illusions, leading to significant false negatives. This work addresses this challenge by formulating it as a perceptual retrieval task and introducing an adaptive view retrieval framework. Built upon a frozen CLIP encoder, the framework dynamically constructs a complementary view bank and hidden-content templates to select reliable views, retrieve potentially harmful content, and calibrate its harmfulness. By moving beyond the limitations of single-view or fixed-transformation strategies, the proposed method achieves a balanced accuracy of 93.2% on the HatefulIllusion benchmark, substantially outperforming existing baselines and matching or exceeding human performance across multiple hallucination datasets.

Abstract

Hateful optical illusions expose a serious gap in current multimodal safety systems. On original-view hateful illusions, previous work shows that six moderation classifiers achieve at most 20.9 to 24.5% accuracy and nine state-of-the-art VLMs remain at or below 10.2% with illusion-aware prompting, leaving most hidden hate undetected. We formulate hidden hateful illusion detection as a perceptual retrieval problem and propose Adaptive View Retrieval. This retrieve-and-calibrate framework assembles a complementary view bank for the image and hidden-message templates, adaptively selects which views to trust, retrieves hidden-message identities, and calibrates whether the recovered evidence is harmful. On HatefulIllusion with a frozen CLIP encoder, Adaptive View Retrieval reaches 93.2% balanced accuracy on the held-out test split. It substantially outperforms original-view baselines and fixed single-transform filters across hate slangs, hate symbols, and visibility levels. The same design also surpasses official fine-tuned CLIP baselines, matches or exceeds human performance on IllusionMNIST, IllusionFashionMNIST, and IllusionAnimals, and outperforms zoom-out preprocessing on HC-Bench under the SemVink protocol. Together, these results show that robust multimodal moderation requires recovering hidden meaning before deciding whether it is harmful.

Problem

• hateful illusions
• multimodal safety
• hidden hate
• content moderation
• perceptual retrieval

Innovation

• Adaptive View Retrieval
• hidden hateful illusions
• multimodal safety
• perceptual retrieval
• CLIP