CVPR Findings2026
A Framework for Evaluating Zero-Shot Image Generation in Concept-Based Explainability
Giacomo Astolfi, Matteo Bianchi, Riccardo Campi, Antonio De Santis, Marco Brambilla
Abstract
Concept-based Explainable Artificial Intelligence (XAI) interprets deep learning models using human-understandable visual features (e.g., textures or object parts) by linking internal representations to class predictions, thereby bridging the gap between low-level image data and high-level semantics. A major challenge, however, is the reliance on large sets of labeled images to represent each concept, which limits scalability. In this work, we investigate the use of zero-shot Text-to-Image (T2I) generative models as a source of synthetic concept datasets for concept-based XAI methods. Specifically, we generate concepts using predefined prompts and evaluate their faithfulness to real ones through four complementary analyses: (1) comparing synthetic vs. real concept images via concept representation similarity; (2) evaluating their intra-similarity by comparing pairs of subsets of the same concept with progressively increasing size; (3) evaluating their performance for downstream explanation tasks using relevant class images; (4) evaluating how removing a concept from tested class images affects explanations of generated concepts. While current T2I generative models promise a shortcut to concept-based XAI, our study highlights challenges and raises open questions about the use of synthetic data generated by zero-shot pipelines in model analyses. The resulting dataset is available at https://github.com/ DataSciencePolimi/ZeroShot-T2I-Concepts.
Ask Lune
Know this paper in minutes, not an afternoon.
Lune can read every page of it and the top-tier papers that cite it, and gives you its claims, its evidence and its limits, each one cited to the paper it comes from. No skimming, no AI slop.
Lineage
- LG-CAV: Train Any Concept Activation Vector with Language Guidance cites Learning Transferable Visual Models From Natural Language Supervision
- Scaling Rectified Flow Transformers for High-Resolution Image Synthesis cites Learning Transferable Visual Models From Natural Language Supervision
Builds on5
- Learning Transferable Visual Models From Natural Language SupervisionAlec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh et al.ICML 2021 · 47,906 citations
- A ConvNet for the 2020sZhuang Liu, Hanzi Mao, Chao-Yuan Wu, Christoph Feichtenhofer et al.CVPR 2022 · 6,782 citations
- Scaling Rectified Flow Transformers for High-Resolution Image SynthesisPatrick Esser, Sumith Kulal, Andreas Blattmann, Rahim Entezari et al.ICML 2024 · 3,620 citations
- LG-CAV: Train Any Concept Activation Vector with Language GuidanceQihan Huang, Jie Song, Mengqi Xue, Haofei Zhang et al.NeurIPS 2024 · 12 citations
- Navigating Neural Space: Revisiting Concept Activation Vectors to Overcome Directional DivergenceFrederik Pahde, Maximilian Dreyer, Moritz Weckbecker, Leander Weber et al.ICLR 2025
Related papers
- Interpreting and Analysing CLIP's Zero-Shot Image Classification via Mutual KnowledgeFawaz Sammani, Nikos DeligiannisNeurIPS 2024 · 15 citations
- Data Descriptions from Large Language Models with Influence EstimationChaeri Kim, Jaeyeon Bae, Taehwan KimEMNLP 2025
- Overlooked Factors in Concept-Based Explanations: Dataset Choice, Concept Learnability, and Human CapabilityVikram V. Ramaswamy, Sunnie S. Y. Kim, Ruth Fong, Olga RussakovskyCVPR 2023
- Can Medical Vision-Language Pre-training Succeed with Purely Synthetic Data?Che Liu, Zhongwei Wan, Haozhe Wang, Yinda Chen et al.ACL Findings 2025 · 3 citations
- ProGen: Progressive Zero-shot Dataset Generation via In-context FeedbackJiacheng Ye, Jiahui Gao, Zhiyong Wu, Jiangtao Feng et al.EMNLP Findings 2022 · 27 citations
