As artificial intelligence systems increasingly rely on multi-source perception and cross-modal learning for human-like visual understanding, their outputs should be both semantically valid and spatially grounded. However, large-scale vision-language benchmarks with precise grounding remain costly to construct, while existing synthetic pipelines often suffer from distribution drift, error propagation, and weak spatial verification. We introduce VQA-G Annotator, a self-improving agentic framework that combines distribution-aware planning, multi-agent orchestration, and spatial reasoning verification in a closed-loop generation, evaluation, and refinement process. Experiments across four benchmark sources show that VQA-G Annotator achieves an average VQAScore of 0.88 for semantic alignment and an Acc@0.5 of 0.57 for spatial grounding, comparing favorably with strong automatic baselines. Downstream and human evaluations further support the utility of the synthesized data 18 for scalable and trustworthy grounded visual understanding.
- Article type
- Year
- Co-author
Open Access
Research Article
Just Accepted
To enhance combat effectiveness and address the challenges of the coordinated scheduling of naval combat equipment, this paper proposes a graph-model-based adaptive maritime kill web generation method.
The proposed method encompasses four key parts. Through battlefield situation modeling, the real-time access and integration of multi-source information are carried out to construct a dynamic battlefield model. A complex task decomposition module is utilized to break down combat tasks into executable subtasks and optimize resource allocation. The kill web is generated based on equipment relationships and capability indicators, and optimized under multiple objective constraints. When the equipment resources change, the kill web is adaptively reconstructed through redundant node supplementation and other means.
Verified by a maritime anti-missile scenario experiment, this method effectively solves the problems of battlefield situation information processing, task decomposition and modeling, and equipment collaborative optimization and dynamic adjustment, and can quickly generate and optimize a kill web to achieve multi-chain and multi-angle defense.
The proposed graph model-based maritime adaptive kill web generation method can improve the overall effectiveness and response ability of modern naval warfare. Future research will continue to optimize the algorithm performance and system response ability to provide stronger support for military operations.
京公网安备11010802044758号