Search papers, labs, and topics across Lattice.
This study introduces a novel methodology for constructing and validating a large-scale multimodal dataset tailored for disaster response, addressing the limitations of existing datasets like Incidents1M and CrisisMMD. By leveraging two Qwen3.5 architectures, the authors generated high-fidelity textual descriptions for 100,000 images and implemented an innovative image-blind LLM-as-a-Judge validation pipeline to ensure semantic alignment for Data-Free Knowledge Distillation (DFKD). The results indicate a high semantic agreement of 78.65/100 between the generated captions and the original images, while revealing critical insights into the precision and recall dynamics of the automated evaluation process.
High-fidelity captions generated for disaster response images reveal a surprising 78.65 semantic agreement, exposing inconsistencies in human annotations.
The deployment of Vision-Language Models (VLMs) in critical domains like disaster management requires high-quality multimodal datasets, especially for transferring knowledge via Data-Free Knowledge Distillation (DFKD). However, existing datasets in this domain either entirely lack descriptive text, such as Incidents1M, or suffer from severe text-image semantic misalignment, such as CrisisMMD. In this work, we present a novel methodology to construct and automatically validate a large-scale multimodal dataset for disaster response. Starting from the vision-only Incidents1M, we successfully recovered 100,000 images and generated high-fidelity textual descriptions using two distinct Qwen3.5 architectures: a 4B dense model and a 35B Mixture-of-Experts (MoE) model. To ensure the generated captions provide reliable semantic anchoring for DFKD, we introduce an image-blind LLM-as-a-Judge validation pipeline leveraging Qwen3.5-9B. By intentionally obscuring the original image from the judge, this evaluator accurately simulates the modality gap of the student model during data-free distillation. Our evaluation across 173,179 label pairs demonstrates a high semantic agreement (78.65/100) between the two architectures. Furthermore, the automated evaluation reveals a conservative captioning behaviour, characterized by a high Precision (77.6%) and low Recall (46.0%). This minimizes the false positive noise, while simultaneously exposing underlying human annotation inconsistencies in the original ground truth. This work provides a scalable, LLM-validated multimodal dataset and a reproducible framework to advance cross-modal knowledge distillation.
