Improving Few-Shot Change Detection Visual Question Answering via Decision-Ambiguity-guided Reinforcement Fine-Tuning
This work addresses a specific bottleneck in CDVQA for remote sensing applications, offering an incremental improvement by optimizing ambiguous samples to enhance model discriminability.
The paper tackled the problem of decision ambiguity in change detection visual question answering (CDVQA), where models struggle with samples having similar confidence between correct answers and distractors, and proposed DARFT, a reinforcement fine-tuning framework that improved performance, achieving consistent gains over supervised fine-tuning baselines, especially in few-shot settings.
Change detection visual question answering (CDVQA) requires answering text queries by reasoning about semantic changes in bi-temporal remote sensing images. A straightforward approach is to boost CDVQA performance with generic vision-language models via supervised fine-tuning (SFT). Despite recent progress, we observe that a significant portion of failures do not stem from clearly incorrect predictions, but from decision ambiguity, where the model assigns similar confidence to the correct answer and strong distractors. To formalize this challenge, we define Decision-Ambiguous Samples (DAS) as instances with a small probability margin between the ground-truth answer and the most competitive alternative. We argue that explicitly optimizing DAS is crucial for improving the discriminability and robustness of CDVQA models. To this end, we propose DARFT, a Decision-Ambiguity-guided Reinforcement Fine-Tuning framework that first mines DAS using an SFT-trained reference policy and then applies group-relative policy optimization on the mined subset. By leveraging multi-sample decoding and intra-group relative advantages, DARFT suppresses strong distractors and sharpens decision boundaries without additional supervision. Extensive experiments demonstrate consistent gains over SFT baselines, particularly under few-shot settings.