Revisiting Change VQA in Remote Sensing with Structured and Native Multimodal Qwen Models

TL;DR

This paper evaluates recent multimodal vision-language models on Change VQA in remote sensing, finding native multimodal models outperform structured pipelines and that larger models do not always yield better results.

Contribution

It provides a comparative analysis of Qwen models on Change VQA, highlighting the effectiveness of native multimodal architectures over structured pipelines.

Findings

Native multimodal models outperform structured vision-language pipelines.

Model size does not monotonically improve performance.

Tightly integrated multimodal backbones are more effective for semantic change reasoning.

Abstract

Change visual question answering (Change VQA) addresses the problem of answering natural-language questions about semantic changes between bi-temporal remote sensing (RS) images. Although vision-language models (VLMs) have recently been studied for temporal RS image understanding, Change VQA remains underexplored in the context of modern multimodal models. In this letter, we revisit the CDVQA benchmark using recent Qwen models under a unified low-rank adaptation (LoRA) setting. We compare Qwen3-VL, which follows a structured vision-language pipeline with multi-depth visual conditioning and a full-attention decoder, with Qwen3.5, a native multimodal model that combines a single-stage alignment with a hybrid decoder backbone. Experimental results on the official CDVQA test splits show that recent VLMs improve over earlier specialized baselines. They further show that performance does not scale monotonically with model size, and that native multimodal models are more effective than structured vision-language pipelines for this task. These findings indicate that tightly integrated multimodal backbones contribute more to performance than scale or explicit multi-depth visual conditioning for language-driven semantic change reasoning in RS imagery.