Seeing Clearly without Training: Mitigating Hallucinations in Multimodal LLMs for Remote Sensing
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2026
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866918367415238656 |
|---|---|
| author | Liu, Yi Zhang, Jing Wang, Di Tian, Xiaoyu Guo, Haonan Du, Bo |
| author_facet | Liu, Yi Zhang, Jing Wang, Di Tian, Xiaoyu Guo, Haonan Du, Bo |
| contents | Multimodal large language models (MLLMs) suffer from pronounced hallucinations in remote sensing visual question-answering (RS-VQA), primarily caused by visual grounding failures in large-scale scenes or misinterpretation of fine-grained small targets. To systematically analyze these issues, we introduce RSHBench, a protocol-based benchmark for fine-grained diagnosis of factual and logical hallucinations. To mitigate grounding-induced factual hallucinations, we further propose Relative Attention-Driven Actively Reasoning (RADAR), a training-free inference method that leverages intrinsic attention in MLLMs to guide progressive localization and fine-grained local reasoning at test time. Extensive experiments across diverse MLLMs demonstrate that RADAR consistently improves RS-VQA performance and reduces both factual and logical hallucinations. Code and data will be publicly available at: https://github.com/MiliLab/RADAR |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2603_02754 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Seeing Clearly without Training: Mitigating Hallucinations in Multimodal LLMs for Remote Sensing Liu, Yi Zhang, Jing Wang, Di Tian, Xiaoyu Guo, Haonan Du, Bo Computer Vision and Pattern Recognition Multimodal large language models (MLLMs) suffer from pronounced hallucinations in remote sensing visual question-answering (RS-VQA), primarily caused by visual grounding failures in large-scale scenes or misinterpretation of fine-grained small targets. To systematically analyze these issues, we introduce RSHBench, a protocol-based benchmark for fine-grained diagnosis of factual and logical hallucinations. To mitigate grounding-induced factual hallucinations, we further propose Relative Attention-Driven Actively Reasoning (RADAR), a training-free inference method that leverages intrinsic attention in MLLMs to guide progressive localization and fine-grained local reasoning at test time. Extensive experiments across diverse MLLMs demonstrate that RADAR consistently improves RS-VQA performance and reduces both factual and logical hallucinations. Code and data will be publicly available at: https://github.com/MiliLab/RADAR |
| title | Seeing Clearly without Training: Mitigating Hallucinations in Multimodal LLMs for Remote Sensing |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2603.02754 |