LandslideAgent with Multimodal LandslideBench: A Domain-Rule-Augmented Agent for Autonomous Landslide Identification and Analysis

2026-06-17

AI Summary

This study addresses the limitations of existing vision-language models in complex geological settings, where capturing fine-grained visual characteristics of landslides while preserving high-level geological semantics remains challenging, often leading to domain hallucinations. To tackle this, the authors introduce LandslideBench, a multimodal fine-grained dataset, and develop LandslideVLM, a specialized fine-tuned model. Furthermore, they propose LandslideAgent, an agent integrating domain-specific rules through a dual-rule controller that leverages structured report metadata and cross-validation constraints to enable knowledge-guided, autonomous tool invocation and end-to-end analysis. Experimental results demonstrate that LandslideVLM outperforms general-purpose models by significant margins—improving landslide detection, fine-grained classification, and semantic description by 10.96%, 32.87%, and 15.91%, respectively.

Abstract

Intelligent landslide hazard interpretation is critical for disaster prevention, yet current paradigms struggle to simultaneously extract visual features and high-level geoscientific semantics, while general-purpose vision-language models (VLMs) suffer from perceptual limitations and domain hallucinations in complex geological scenarios. To address these challenges, we propose an instruction-driven agentic framework comprising three components. First, LandslideBench, a multimodal fine-grained dataset with seven subtype labels, high-resolution imagery, pixel-level masks, and high-quality textual descriptions, is constructed via multi-VLM cross-validation and interactive annotation. Then, LandslideVLM, a landslide-oriented VLM, is fine-tuned via LoRA on LandslideBench to enhance geological semantic understanding. Finally, LandslideAgent, a domain rule-enhanced agent taking LandslideVLM as its cognitive backbone, employs a dual-rule controller incorporating structured report metadata constraints and cross-validation identification constraints to regulate automated tool invocation. Experiments demonstrate that LandslideBench provides effective baselines across five mainstream models on fine-grained classification and semantic segmentation. LandslideVLM achieves accuracy improvements of 10.96%, 32.87%, and 15.91% on landslide discrimination, fine-grained classification, and semantic description quality, respectively. LandslideAgent further enables autonomous multi-source spatial data inference, realizing full-process intelligence for landslide identification and analysis.

Problem

• landslide identification
• vision-language models
• geoscientific semantics
• domain hallucination
• multimodal perception

Innovation

• LandslideAgent
• LandslideBench
• domain-rule-augmented
• vision-language model
• autonomous landslide analysis