{"ID":6138202,"CreatedAt":"2026-07-09T01:07:32.349475501Z","UpdatedAt":"2026-07-11T10:38:06.360441493Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2607.07264","arxiv_id":"2607.07264","title":"Naming the Concepts Classifiers Rely On: Language-Anchored Decomposition for Faithful Explanation","abstract":"Deep neural networks are widely deployed in high-stakes visual applications where interpretability is critical, yet existing explanations face a trade-off: post-hoc concept methods recover factors that are faithful to a model's behavior but unnamed, while naming and by-design methods attach human-readable concepts only by retraining or altering the classifier. We propose Language-Anchored Decomposition (LAD), a post-hoc framework that delivers concepts which are simultaneously named, faithful, and obtained without modifying the model. For each class, a large language model proposes a concept vocabulary that CLIP-based similarity maps localize across image regions. Inverting standard non-negative matrix factorization, LAD fixes these language-grounded maps as the coefficient matrix and learns only a concept basis that reconstructs the frozen encoder's activations, so naming becomes a structural constraint and the model's own feature geometry determines which concepts are retained. Removing this anchor preserves accuracy but collapses attribution faithfulness. Across natural-image, scene, and medical-imaging benchmarks, LAD produces spatially precise explanations that are decision-relevant under both concept insertion and deletion, while uniquely providing stable, human-interpretable concept names.","short_abstract":"Deep neural networks are widely deployed in high-stakes visual applications where interpretability is critical, yet existing explanations face a trade-off: post-hoc concept methods recover factors that are faithful to a model's behavior but unnamed, while naming and by-design methods attach human-readable concepts only...","url_abs":"https://arxiv.org/abs/2607.07264","url_pdf":"https://arxiv.org/pdf/2607.07264v1","authors":"[\"Ahsan Habib Akash\",\"Dipkamal Bhusal\",\"Stacey Jones\",\"Donald A. Adjeroh\",\"Binod Bhattarai\",\"Prashnna Kumar Gyawali\"]","published":"2026-07-08T10:50:04Z","proceeding":"cs.CV","tasks":"[\"cs.CV\"]","methods":"[\"Language Model\"]","has_code":false}