New multimodal dataset Theia generated for disaster response using Qwen3.5

Researchers have developed a new methodology to create and validate a large-scale multimodal dataset for disaster response, named Theia. This dataset is derived from the vision-only Incidents1M dataset and features high-fidelity textual descriptions generated using two Qwen3.5 model architectures (4B and 35B MoE). An automated validation pipeline, employing an image-blind LLM-as-a-Judge approach with Qwen3.5-9B, was used to ensure semantic accuracy and assess captioning behavior, revealing a high precision but low recall. AI

IMPACT This work provides a scalable, LLM-validated multimodal dataset and a reproducible framework to advance cross-modal knowledge distillation for critical domains like disaster response.

RANK_REASON The cluster describes a research paper detailing the creation and validation of a new multimodal dataset. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CV →

• Adriano Mancini
• CrisisMMD
• Incidents1M: a Large-Scale Dataset of Images With Natural Disasters, Damage, and Incidents
• LLM-as-a-Judge
• Qwen 3.5
• Qwen3.5 35B
• Qwen3.5 4B
• Qwen3.5:9b
• Theia

• paper
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →