{"id":"arxiv:2604.04575","kind":"preprint","title":"Erasure or Erosion? Evaluating Compositional Degradation in Unlearned Text-To-Image Diffusion Models","authors":["Arian Komaei Koma","Seyed Amir Kasaei","Ali Aghayari","AmirMahdi Sadeghzadeh","Mohammad Hossein Rohban"],"author_count":5,"year":2026,"published":"2026-04-06","venue":null,"cited_by":null,"doi":null,"arxiv_id":"2604.04575","nct_id":null,"source_url":"https://arxiv.org/abs/2604.04575","archive_url":"https://space-kitty.com/papers/arxiv:2604.04575","updated":"2026-04-06","pmid":null,"pmcid":null,"openalex_id":null,"pdf_url":"https://arxiv.org/pdf/2604.04575","abstract":"Post-hoc unlearning has emerged as a practical mechanism for removing undesirable concepts from large text-to-image diffusion models. However, prior work primarily evaluates unlearning through erasure success; its impact on broader generative capabilities remains poorly understood. In this work, we conduct a systematic empirical study of concept unlearning through the lens of compositional text-to-image generation. Focusing on nudity removal in Stable Diffusion 1.4, we evaluate a diverse set of state-of-the-art unlearning methods using T2I-CompBench++ and GenEval, alongside established unlearning benchmarks. Our results reveal a consistent trade-off between unlearning effectiveness and compositional integrity: methods that achieve strong erasure frequently incur substantial degradation in attribute binding, spatial reasoning, and counting. Conversely, approaches that preserve compositional structure often fail to provide robust erasure. These findings highlight limitations of current evaluation practices and underscore the need for unlearning objectives that explicitly account for semantic preservation beyond targeted suppression.","abstract_license":"CC0-1.0 (arXiv descriptive metadata)","license":null,"is_open_access":false,"retracted":false,"peer_reviewed":false,"categories":["cs.CV"],"details":null,"provenance":"arxiv","attribution":"Abstract: arXiv descriptive metadata (CC0). Thank you to arXiv for use of its open access interoperability. This service was not reviewed or approved by, nor does it necessarily express or reflect the policies or opinions of, arXiv.","topics":[],"cited_in_dossiers":[]}