UNBOX: Unveiling Black-box visual models with Natural-language

2026-03-09

AI Summary

This work proposes UNBOX, a novel approach to interpreting visual models under strict black-box conditions—without access to training data, gradients, or internal model parameters. By leveraging large language models and text-to-image diffusion models, UNBOX reformulates activation maximization as a semantic search problem grounded in output probabilities, generating human-interpretable textual descriptions that reveal the concepts learned by the model. Unlike prior methods requiring white- or gray-box assumptions, UNBOX enables class-level semantic dissection of vision models solely through their API outputs. Experiments on ImageNet-1K, Waterbirds, and CelebA demonstrate that UNBOX achieves semantic fidelity, visual feature relevance, and bias slice discovery comparable to white-box techniques, establishing that high-value model insights remain attainable even in fully black-box settings.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Natural Language Processing: Interpretability, Analysis, and Evaluation of NLP Models
• Machine Learning: Large Multimodal Models (LMMs)

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

Ensuring trustworthiness in open-world visual recognition requires models that are interpretable, fair, and robust to distribution shifts. Yet modern vision systems are increasingly deployed as proprietary black-box APIs, exposing only output probabilities and hiding architecture, parameters, gradients, and training data. This opacity prevents meaningful auditing, bias detection, and failure analysis. Existing explanation methods assume white- or gray-box access or knowledge of the training distribution, making them unusable in these real-world settings. We introduce UNBOX, a framework for class-wise model dissection under fully data-free, gradient-free, and backpropagation-free constraints. UNBOX leverages Large Language Models and text-to-image diffusion models to recast activation maximization as a purely semantic search driven by output probabilities. The method produces human-interpretable text descriptors that maximally activate each class, revealing the concepts a model has implicitly learned, the training distribution it reflects, and potential sources of bias. We evaluate UNBOX on ImageNet-1K, Waterbirds, and CelebA through semantic fidelity tests, visual-feature correlation analyses and slice-discovery auditing. Despite operating under the strictest black-box constraints, UNBOX performs competitively with state-of-the-art white-box interpretability methods. This demonstrates that meaningful insight into a model's internal reasoning can be recovered without any internal access, enabling more trustworthy and accountable visual recognition systems.

Problem

• black-box models
• model interpretability
• bias detection
• visual recognition
• trustworthiness

Innovation

• black-box interpretability
• activation maximization
• large language models
• diffusion models
• model auditing