This work presents a comparative analysis of embedding-based and generative models for classifying geoscience technical documents. Using a multi-disciplinary benchmark dataset, we evaluated the trade-offs between model accuracy, stability, and computational cost. We find that generative Vision-Language Models (VLMs) like Qwen2.5-VL, enhanced with Chain-of-Thought (CoT) prompting, achieve superior zero-shot accuracy (82%) compared to state-of-the-art multimodal embedding models like QQMM (63%). We also demonstrate that while supervised fine-tuning (SFT) can improve VLM performance, it is sensitive to training data imbalance.
@article{arxiv.2604.04997,
title = {Evaluation of Embedding-Based and Generative Methods for LLM-Driven Document Classification: Opportunities and Challenges},
author = {Rong Lu and Hao Liu and Song Hou},
journal= {arXiv preprint arXiv:2604.04997},
year = {2026}
}
Comments
Accepted at the IMAGE'25 Workshop (PCW-11), Society of Exploration Geophysicists (SEG). Published version available at https://doi.org/10.1190/image2025-w11-03.1