HorusEye: Language as Dynamic Attention for Emergency Visual Analysis
Not on the map. Why?
Abstract (arXiv)
We introduce HorusEye, Language as Dynamic Attention for Emergency Visual Analysis. Our investigation followed five stages. The first one is benchmarking RefCOCO-Degraded, a dataset of 15,244 images (3,811 base images x 4 conditions: Clean, Fog, Smoke and Thermal) with systematic visual degradation. Through four research questions, we evaluate multiple VLMs (Gemini, Qwen2-VL, BLIP-2, LLaVA, Kosmos-2) across visual grounding the second stage, language feedback recovery the third one, health VQA tasks the fourth, and hallucination analysis the final stage. Our key finding is that language feedback effectiveness is model-dependent: Gemini achieves +47.3% improvement in thermal conditions through iterative language feedback, while Qwen2-VL shows -5.1% degradation under the same protocol. We also identify the 'Thermal Paradox' where cropping strategies that improve RGB performance catastrophically fail in thermal imagery. Furthermore, BLIP-2 uniquely hallucinates more under degradation, making it unsuitable for emergency deployment
Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 20 of its 26 references that are in the map. First posted on arXiv 2 Jun 2026.
Builds on
20 of 26 references.
References · 20
- 2014Microsoft COCO: Common Objects in Contextcited by 17,608
- 2023Visual Instruction Tuningcited by 7,869
- 2023BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Modelscited by 6,137
- 2023Improved Baselines with Visual Instruction Tuningcited by 4,418
- 2024Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolutioncited by 4,234
- 2024Gemini 1.5: Unlocking multimodal understanding across millions of tokens of contextcited by 3,224
- 2019Benchmarking Neural Network Robustness to Common Corruptions and Perturbationscited by 2,877
- 2016Making the V in VQA Matter: Elevating the Role of Image Understanding in Visual Question Answeringcited by 2,516
- 2015VQA: Visual Question Answeringcited by 2,506
- 2013Rich feature hierarchies for accurate object detection and semantic segmentationcited by 2,162
- 2023Evaluating Object Hallucination in Large Vision-Language Modelscited by 1,729
- 2015Show, Attend and Tell: Neural Image Caption Generation with Visual Attentioncited by 1,423
8 more
- 2017Bottom-Up and Top-Down Attention for Image Captioning and Visual Question Answeringcited by 1,312
- 2023Kosmos-2: Grounding Multimodal Large Language Models to the Worldcited by 1,089
- 2016Modeling Context in Referring Expressionscited by 1,063
- 2018Object Hallucination in Image Captioningcited by 669
- 2017Benchmarking Single Image Dehazing and Beyondcited by 447
- 2017Semantic Foggy Scene Understanding with Synthetic Datacited by 403
- 2021TransVG: End-to-End Visual Grounding with Transformerscited by 271
- 2024THRONE: An Object-based Hallucination Benchmark for the Free-form Generations of Large Vision-Language Modelscited by 36
Built on by
Not cited yet.
Similar, not cited
Close in wording, no citation either way
- 2025Better Safe Than Sorry? Overreaction Problem of Vision Language Models in Visual Emergency Recognition
- 2026ThermEval: A Structured Benchmark for Evaluation of Vision-Language Models on Thermal Imagery
- 2025HoliSafe: Holistic Safety Benchmarking and Modeling for Vision-Language Model
- 2026Are Vision-Language-Action Models Robust to One-Step Observation Perturbations?
- 2024BEAF: Observing BEfore-AFter Changes to Evaluate Hallucination in Vision-language Models
- 2023Lost in Translation: When GPT-4V(ision) Can't See Eye to Eye with Text. A Vision-Language-Consistency Analysis of VLLMs and Beyond
- 2026EviSafe: Evidence-Grounded Safety Evaluation for Vision-Language Models
- 2025VSCBench: Bridging the Gap in Vision-Language Model Safety Calibration