Falcon: Functional Assembly and Language for Compositional Reasoning in X-ray

2026-06-24

AI Summary

This work addresses a critical limitation of current vision-language models, which are object-centric and struggle to recognize composite threats in X-ray security screening—threats composed of spatially dispersed yet functionally compatible components such as batteries, detonators, and explosives. The study formalizes, for the first time, the task of “composite threat reasoning” and introduces a multimodal framework that leverages segmentation-aware feature extraction to construct a structured safety state representation encoding component presence, pairwise functional compatibility, and scene-level risk. This representation serves as an explicit intermediate interface injected into a language model, enabling relationally consistent and safety-aware reasoning. Evaluated on the newly introduced Falcon-X benchmark, the proposed method significantly improves component localization accuracy and threat assessment consistency, establishing composite safety reasoning as a novel evaluation paradigm for multimodal systems.

Abstract

Conventional vision-language models are largely object-centric, focusing on detecting and describing individual entities. In safety-critical X-ray baggage screening, however, threat often emerges not from a single object but from the functional compatibility of spatially dispersed components, such as batteries, detonators, and explosive charges. We formalize this setting as \emph{compositional threat reasoning}, where risk is modeled as a relational property of grounded regions rather than an independent detection outcome. We introduce \textbf{Falcon}, a multimodal framework that abstracts segmentation-aware region features into a structured safety state capturing component presence, pairwise functional compatibility, and scene-level risk. This structured representation is injected into the language model as an explicit intermediate interface, encouraging relationally consistent and safety-aware reasoning. To evaluate this problem, we present \textbf{Falcon-X}, a benchmark that unifies dense grounding with structured supervision over component completeness and risk inference in cluttered X-ray imagery. Experiments show that while existing multimodal models adapt to appearance, they struggle with compositional safety reasoning. Falcon improves functional grounding and produces more coherent threat assessments, establishing compositional safety reasoning as a distinct evaluation paradigm for multimodal systems.

Problem

• compositional threat reasoning
• X-ray baggage screening
• functional compatibility
• multimodal reasoning
• safety-critical perception

Innovation

• compositional threat reasoning
• functional compatibility
• structured safety state
• multimodal reasoning
• X-ray baggage screening