Falcon: A Cross-Modal Evaluation Dataset for Comprehensive Safety Perception

• URL: http://arxiv.org/abs/2509.23783v1
• Date: Sun, 28 Sep 2025 10:00:37 GMT
• Title: Falcon: A Cross-Modal Evaluation Dataset for Comprehensive Safety Perception
• Abstract summary: This work highlights the crucial role of visual information in moderating content in visual question answering.<n>We introduce Falcon, a large-scale vision-language safety dataset containing 57,515 VQA pairs across 13 harm categories.<n>We propose FalconEye, a specialized evaluator fine-tuned from Qwen2.5-VL-7B using the Falcon dataset.
• Score: 19.056814278422383
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Existing methods for evaluating the harmfulness of content generated by large language models (LLMs) have been well studied. However, approaches tailored to multimodal large language models (MLLMs) remain underdeveloped and lack depth. This work highlights the crucial role of visual information in moderating content in visual question answering (VQA), a dimension often overlooked in current research. To bridge this gap, we introduce Falcon, a large-scale vision-language safety dataset containing 57,515 VQA pairs across 13 harm categories. The dataset provides explicit annotations for harmful attributes across images, instructions, and responses, thereby facilitating a comprehensive evaluation of the content generated by MLLMs. In addition, it includes the relevant harm categories along with explanations supporting the corresponding judgments. We further propose FalconEye, a specialized evaluator fine-tuned from Qwen2.5-VL-7B using the Falcon dataset. Experimental results demonstrate that FalconEye reliably identifies harmful content in complex and safety-critical multimodal dialogue scenarios. It outperforms all other baselines in overall accuracy across our proposed Falcon-test dataset and two widely-used benchmarks-VLGuard and Beavertail-V, underscoring its potential as a practical safety auditing tool for MLLMs.

Related papers

• SafeAtlas-VL: Beyond Binary Multimodal Safety with Large-Scale Data and Guard Models [73.04773649437375]
  Multimodal safety moderation requires distinguishing risks arising from visual content, user intent, and assistant behavior.<n>We introduce SafeAtlas-VL, a dataset of 1.5M training instances that places image-, request-, and response-level judgments on a five-level ordered scale.<n>We also construct SafeAtlas-Bench, a set of 5,000 instances for evaluating five-level predictions and continuous risk scores.
  arXiv  Detail & Related papers  (2026-08-29T07:08:01Z)
• Benchmarking Large Vision-Language Models on Fine-Grained Image Tasks: From Evaluation to Diagnosis [63.49939245944781]
  We introduce FG-BMK, a comprehensive fine-grained evaluation benchmark containing 1.01 million questions and 0.28 million images.<n>We find that current LVLMs remain inadequate fine-grained recognizers, with failures arising from intertwined bottlenecks in visual representations, semantic grounding, modality alignment, and category-level knowledge.
  arXiv  Detail & Related papers  (2026-06-17T13:23:26Z)
• OutSafe-Bench: A Benchmark for Multimodal Offensive Content Detection in Large Language Models [54.80460603255789]
  We introduce OutSafe-Bench, the first most comprehensive content safety evaluation test suite designed for the multimodal era.<n>OutSafe-Bench includes a large-scale dataset that spans four modalities, featuring over 18,000 bilingual (Chinese and English) text prompts, 4,500 images, 450 audio clips and 450 videos, all systematically annotated across nine critical content risk categories.<n>In addition to the dataset, we introduce a Multidimensional Cross Risk Score (MCRS), a novel metric designed to model and assess overlapping and correlated content risks across different categories.
  arXiv  Detail & Related papers  (2025-11-13T13:18:27Z)
• Visual-Semantic Knowledge Conflicts in Operating Rooms: Synthetic Data Curation for Surgical Risk Perception in Multimodal Large Language Models [7.916129615051081]
  We introduce a dataset comprising over 34,000 synthetic images generated by diffusion models.<n>The dataset includes 214 human-annotated images that serve as a gold-standard reference for validation.
  arXiv  Detail & Related papers  (2025-06-25T07:06:29Z)
• HoliSafe: Holistic Safety Benchmarking and Modeling for Vision-Language Model [58.12612140992874]
  We introduce a holistic safety dataset and benchmark, textbfHoliSafe, that spans all five safe/unsafe image-text combinations.<n>We also propose a novel modular framework for enhancing VLM safety with a visual guard module (VGM) designed to assess the harmfulness of input images.<n> Experiments show that Safe-VLM with VGM, trained on our HoliSafe, achieves state-of-the-art safety performance across multiple VLM benchmarks.
  arXiv  Detail & Related papers  (2025-06-05T07:26:34Z)
• Unlearning Sensitive Information in Multimodal LLMs: Benchmark and Attack-Defense Evaluation [88.78166077081912]
  We introduce a multimodal unlearning benchmark, UnLOK-VQA, and an attack-and-defense framework to evaluate methods for deleting specific multimodal knowledge from MLLMs.<n>Our results show multimodal attacks outperform text- or image-only ones, and that the most effective defense removes answer information from internal model states.
  arXiv  Detail & Related papers  (2025-05-01T01:54:00Z)
• SafeBench: A Safety Evaluation Framework for Multimodal Large Language Models [75.67623347512368]
  We propose toolns, a comprehensive framework designed for conducting safety evaluations of MLLMs. Our framework consists of a comprehensive harmful query dataset and an automated evaluation protocol. Based on our framework, we conducted large-scale experiments on 15 widely-used open-source MLLMs and 6 commercial MLLMs.
  arXiv  Detail & Related papers  (2024-10-24T17:14:40Z)
• T2Vs Meet VLMs: A Scalable Multimodal Dataset for Visual Harmfulness Recognition [24.78672820633581]
  Existing harmful datasets are curated by the presence of a narrow range of harmful objects. This hinders the generalizability of methods based on such datasets, potentially leading to misjudgments. We propose a comprehensive harmful dataset, consisting of 10,000 images and 1,000 videos, crawled from the Internet and generated by 4 generative models.
  arXiv  Detail & Related papers  (2024-09-29T15:20:00Z)
• FineCops-Ref: A new Dataset and Task for Fine-Grained Compositional Referring Expression Comprehension [10.482908189805872]
  Referring Expression (REC) is a crucial cross-modal task that objectively evaluates the capabilities of language understanding, image comprehension, and language-to-image grounding.<n>We have established a new REC dataset characterized by two key features.<n>It includes negative text and images created through fine-grained editing and generation based on existing data.
  arXiv  Detail & Related papers  (2024-09-23T06:56:51Z)
• SPA-VL: A Comprehensive Safety Preference Alignment Dataset for Vision Language Model [77.86593720792986]
  We propose a Safety Preference Alignment dataset for Vision Language Models named SPA-VL.<n> SPA-VL covers 6 harmfulness domains, 13 categories, and 53 subcategories, and contains 100,788 samples of the quadruple (question, image, chosen response, rejected response)<n>Experiments indicate that models trained with alignment techniques on the SPA-VL dataset exhibit substantial improvements in harmlessness and helpfulness while maintaining core capabilities.
  arXiv  Detail & Related papers  (2024-06-17T18:57:37Z)
• MLLMGuard: A Multi-dimensional Safety Evaluation Suite for Multimodal Large Language Models [39.97454990633856]
  We present MLLMGuard, a multidimensional safety evaluation suite for MLLMs. It includes a bilingual image-text evaluation dataset, inference utilities, and a lightweight evaluator. Our evaluation results across 13 advanced models indicate that MLLMs still have a substantial journey ahead before they can be considered safe and responsible.
  arXiv  Detail & Related papers  (2024-06-11T13:41:33Z)
• VALOR-EVAL: Holistic Coverage and Faithfulness Evaluation of Large Vision-Language Models [57.43276586087863]
  Large Vision-Language Models (LVLMs) suffer from hallucination issues, wherein the models generate plausible-sounding but factually incorrect outputs. Existing benchmarks are often limited in scope, focusing mainly on object hallucinations. We introduce a multi-dimensional benchmark covering objects, attributes, and relations, with challenging images selected based on associative biases.
  arXiv  Detail & Related papers  (2024-04-22T04:49:22Z)
• B-AVIBench: Towards Evaluating the Robustness of Large Vision-Language Model on Black-box Adversarial Visual-Instructions [73.97665608366447]
  Large Vision-Language Models (LVLMs) have shown significant progress in responding well to visual-instructions from users.<n>These instructions, encompassing images and text, are susceptible to both intentional and inadvertent attacks.<n>We introduce B-AVIBench, a framework designed to analyze the robustness of LVLMs when facing various Black-box Adrial Visual-Instructions.
  arXiv  Detail & Related papers  (2024-03-14T12:51:07Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.