Low-cost concept-based localized explanations: How far can we get with training-free approaches?

2026-06-27

AI Summary

This work addresses the scarcity of concept annotations that hinders the validation of concept-based explainable AI (C-XAI) at fine-grained local image regions. It investigates the capability of medium-scale multimodal large language models (MLLMs, 7B–32B parameters) to name concepts in localized image areas—spanning both object and part levels—under zero-shot, training-free conditions. The study introduces two reproducible evaluation protocols, CoNa and Open-CoNa, tailored respectively for moderate-sized vocabularies and large-scale label spaces, leveraging class-constrained prompting and embedding similarity matching for efficient annotation. Experiments across multiple datasets achieve object-level exact-match accuracies ranging from 62% to 88%, demonstrating for the first time the feasibility and effectiveness of training-free approaches for local concept explanation.

Abstract

Concept-based Explainable AI (C-XAI) seeks human-understandable explanations grounded in semantic concepts, yet validation is limited by the scarcity of fine-grained concept annotations. We evaluate whether mid-scale Multimodal Large Language Models (MLLMs) can perform localized concept naming under strict zero-shot conditions by assigning labels to bounding-box regions at both object and part levels. We propose a reproducible zero-shot evaluation protocol for Concept Naming (CoNa) with (i) closed-set, category-constrained prompting for moderate vocabularies and (ii) Open-CoNa, an embedding-similarity-based strategy for large label spaces. Experiments with four MLLMs (7B-32B) show consistent performance trends across datasets, reaching 62%-88% object-level exact-match accuracy, highlighting the potential of training-free concept annotation from localized regions. We discuss limitations and failure modes and release a reproducible framework to support future low-cost C-XAI research.

Problem

• Concept-based Explainable AI
• zero-shot concept naming
• localized explanations
• multimodal large language models
• fine-grained concept annotations

Innovation

• Concept-based Explainable AI
• zero-shot concept naming
• Multimodal Large Language Models
• training-free explanation
• localized concept annotation