LU-500: A Logo Benchmark for Concept Unlearning

• URL: http://arxiv.org/abs/2607.24101v2
• Date: Wed, 29 Jul 2026 12:52:15 GMT
• Title: LU-500: A Logo Benchmark for Concept Unlearning
• Abstract summary: Concept unlearning is increasingly used to limit the reproduction of protected or unsafe visual concepts in text-to-image models.<n>We introduce LU-500, a logo-unlearning benchmark built from Fortune Global 500 companies to study this localized and semantically entangled setting.<n>To avoid reducing the task to a binary detector score, we define a multi-grained protocol that evaluates both local logo removal and global image preservation.
• Score: 27.279391309749624
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Concept unlearning is increasingly used to limit the reproduction of protected or unsafe visual concepts in text-to-image models. Existing evaluations, however, mostly study targets that dominate the whole image, such as styles, broad object categories, or portrait-like identities, leaving company logos comparatively underexamined. Logos create a different failure mode: a small localized mark can carry the entire protected concept, must be visually precise to remain recognizable, and can be triggered implicitly by products, storefronts, packaging, or advertisements even when the word ``logo'' is absent. We introduce LU-500, a logo-unlearning benchmark built from Fortune Global 500 companies to study this localized and semantically entangled setting. LU-500 contains nearly 10,000 curated text-query and logo-image pairs, with an explicit track (LUex-500) and an implicit contextual track (LUim-500). To avoid reducing the task to a binary detector score, we define a multi-grained protocol that evaluates both local logo removal and global image preservation in pixel and latent spaces. Experiments on representative inference-time methods, including NP, SLD, and SEGA, and compatible fine-tuning-based methods such as ESD and Forget-Me-Not, show that the evaluated methods struggle to remove logo evidence without changing non-target content. We further analyze ProLU, a prompt-space multi-agent baseline: it improves local erasure by removing logo-inducing semantics, but also illustrates why prompt filtering is not a substitute for weight-level disentanglement. Correlation analyses over logo area, location, and structural complexity suggest that future logo unlearning may need spatially aware controls, such as SSIM-guided constraints, rather than purely global concept suppression.

Related papers

• From Unlearning to UNBRANDING: A Benchmark for Trademark-Safe Text-to-Image Generation [0.7798283447125206]
  Brand recognition is multi-dimensional, extending beyond explicit logos to encompass distinctive structural features.<n>We introduce unbranding, a novel task for the fine-grained removal of both trademarks and subtle structural brand features.<n>Our results, validated by our Vision Language Models metric, confirm unbranding is a distinct, practically relevant problem.
  arXiv  Detail & Related papers  (2025-12-15T23:15:36Z)
• Introducing Visual Scenes and Reasoning: A More Realistic Benchmark for Spoken Language Understanding [51.010563573083495]
  We introduce VRSLU, a novel dataset that integrates both Visual images and explicit Reasoning.<n>For over-idealized CA, we use GPT-4o and FLUX.1-dev to generate images reflecting users' environments and statuses, followed by human verification to ensure quality.<n>For reasoning, GPT-4o is employed to generate explanations for predicted labels, which are then refined by human annotators to ensure accuracy and coherence.
  arXiv  Detail & Related papers  (2025-11-24T11:32:24Z)
• Logo-VGR: Visual Grounded Reasoning for Open-world Logo Recognition [25.658499211854153]
  We introduce an open-world logo recognition benchmark, a core challenge in product moderation.<n>Unlike traditional logo recognition methods that rely on memorizing representations of tens of thousands of brands, we propose Logo-VGR.<n>We show that Logo-VGR outperforms strong baselines by nearly 10 points in OOD settings.
  arXiv  Detail & Related papers  (2025-09-30T05:35:10Z)
• LogoSticker: Inserting Logos into Diffusion Models for Customized Generation [73.59571559978278]
  We introduce the task of logo insertion into text-to-image models. Our goal is to insert logo identities into diffusion models and enable their seamless synthesis in varied contexts. We present a novel two-phase pipeline LogoSticker to tackle this task.
  arXiv  Detail & Related papers  (2024-07-18T17:54:49Z)
• SLANT: Spurious Logo ANalysis Toolkit [61.59021920232986]
  We develop SLANT: A Spurious Logo ANalysis Toolkit. It contains a semi-automatic mechanism for mining such "spurious" logos. We uncover various seemingly harmless logos that VL models correlate with negative human adjectives. An attacker could place a spurious logo on harmful content, causing the model to misclassify it as harmless.
  arXiv  Detail & Related papers  (2024-06-03T15:41:31Z)
• Image-Text Pre-Training for Logo Recognition [0.27195102129094995]
  We propose two novel contributions to improve the matching model's performance. A standard paradigm of fine-tuning ImageNet pre-trained models fails to discover the text sensitivity necessary to solve the matching problem effectively. We show that the same vision backbone pre-trained on image-text data, when fine-tuned on OpenLogoDet3K47, achieves $98.6%$ recall@1.
  arXiv  Detail & Related papers  (2023-09-18T23:18:02Z)
• Contrastive Multi-View Textual-Visual Encoding: Towards One Hundred Thousand-Scale One-Shot Logo Identification [2.243832625209014]
  We study the problem of identifying logos of business brands in natural scenes in an open-set one-shot setting. We propose a novel multi-view textual-visual encoding framework that encodes text appearing in the logos. We evaluate our proposed framework for cropped logo verification, cropped logo identification, and end-to-end logo identification in natural scene tasks.
  arXiv  Detail & Related papers  (2022-11-23T12:59:41Z)
• Discriminative Semantic Feature Pyramid Network with Guided Anchoring for Logo Detection [52.36825190893928]
  We propose a novel approach, named Discriminative Semantic Feature Pyramid Network with Guided Anchoring (DSFP-GA) Our approach mainly consists of Discriminative Semantic Feature Pyramid (DSFP) and Guided Anchoring (GA)
  arXiv  Detail & Related papers  (2021-08-31T11:59:00Z)
• An Effective and Robust Detector for Logo Detection [58.448716977297565]
  Some attackers fool the well-trained logo detection model for infringement. A novel logo detector based on the mechanism of looking and thinking twice is proposed in this paper. We extend detectoRS algorithm to a cascade schema with an equalization loss function, multi-scale transformations, and adversarial data augmentation.
  arXiv  Detail & Related papers  (2021-08-01T10:17:53Z)
• LogoDet-3K: A Large-Scale Image Dataset for Logo Detection [61.296935298332606]
  We introduce LogoDet-3K, the largest logo detection dataset with full annotation. It has 3,000 logo categories, about 200,000 manually annotated logo objects and 158,652 images. We propose a strong baseline method Logo-Yolo, which incorporates Focal loss and CIoU loss into the state-of-the-art YOLOv3 framework for large-scale logo detection.
  arXiv  Detail & Related papers  (2020-08-12T14:57:53Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.