{"total":18,"items":[{"citing_arxiv_id":"2606.05290","ref_index":8,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Do Models Share Safety Representations? Cross-Model Steering for Safe Visual Generation","primary_cat":"cs.CV","submitted_at":"2026-06-03T18:00:04+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":8.0,"formal_verification":"none","one_line_summary":"A safety direction estimated in a source LLM is transported to a target generator through lightweight alignment on benign data alone, matching native safety performance without any target-side unsafe data.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.29230","ref_index":3,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Toward Ethical Facial Age Estimation: A Generalized Zero-Shot Benchmark Without Training on Children's Data","primary_cat":"cs.CV","submitted_at":"2026-05-28T01:44:39+00:00","verdict":"CONDITIONAL","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"A generalized zero-shot benchmark is introduced for facial age estimation that excludes all children's data from training and demonstrates consistent failure of nine state-of-the-art methods to generalize to unseen young age groups.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.28137","ref_index":2,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"No Safe Dose: How Training Data Drives Unsafe Image Generation","primary_cat":"cs.CV","submitted_at":"2026-05-27T08:21:38+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Proportion of unsafe images in training data directly increases unsafe outputs in text-to-image models, independent of absolute count, with complementary risk reduction from safer text encoders.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.22098","ref_index":4,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"TextTeacher: What Can Language Teach About Images?","primary_cat":"cs.CV","submitted_at":"2026-05-21T07:36:00+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"TextTeacher uses frozen text embeddings from captions as semantic anchors to guide vision model training, improving ImageNet accuracy by up to 2.7 p.p. and transfer performance by 1.0 p.p. on average.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2604.16483","ref_index":4,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Dynamic Eraser for Guided Concept Erasure in Diffusion Models","primary_cat":"cs.CV","submitted_at":"2026-04-13T02:25:14+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"DSS is a lightweight inference-time framework that erases concepts in diffusion models at 91% average rate while preserving image fidelity, outperforming prior methods.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2602.04759","ref_index":13,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"How to Stop Playing Whack-a-Mole: Mapping the Ecosystem of Technologies Facilitating AI-Generated Non-Consensual Intimate Images","primary_cat":"cs.CY","submitted_at":"2026-02-04T16:58:05+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"The paper introduces the first comprehensive taxonomy and visualization of 11 categories of technologies facilitating AI-generated non-consensual intimate images, derived from synthesis of primary sources and demonstrated through case studies.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2602.00033","ref_index":12,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Mapping the Stochastic Penal Colony","primary_cat":"cs.CY","submitted_at":"2026-01-18T16:22:59+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":4.0,"formal_verification":"none","one_line_summary":"Content moderation operates as a stochastic penal colony that banishes users through the constant threat of account suspension, shown via auto-ethnographic case studies of Twitter, OpenAI DALL-E 2, and Pinterest.","context_count":1,"top_context_role":"background","top_context_polarity":"support","context_text":"The system card [37, 80] for the original release in April 2022 makes clear thatdall•e2 was, in part, trained with \"publicly available sources\" [64], which in all likelihood included Internet-sourced data similar to the LAION-400M dataset [91]. While OpenAI has declined to elaborate on the exact sources fordall•e's training data, we know that such Internet-sourced datasets are anything but safe [12]. That makesdall•eunsafe by design. As shown in Appendix B.1 on page 21,dall•e's content policy came right out against anything \"that could cause harm. \" In theirfaqentry fordall•e's warnings [70], OpenAI claimed that \"safe usage of the platform is our highest priority. \" Yetdall•e's content policy disallowedallpolitical and health content. Those prohibitions are not just unusu-"},{"citing_arxiv_id":"2509.11487","ref_index":3,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Collective Recourse for Generative Urban Visualizations","primary_cat":"cs.HC","submitted_at":"2025-09-15T00:39:59+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"Collective recourse formalizes community reports to fix group harms in diffusion models for urban visualizations via a report-triage-fix-verify pipeline, four primitives, a mandate score, and synthetic evaluation of 240 reports.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2506.17185","ref_index":16,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"A Common Pool of Privacy Problems: Legal and Technical Lessons from a Large-Scale Web-Scraped Machine Learning Dataset","primary_cat":"cs.CR","submitted_at":"2025-06-20T17:40:05+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"An empirical audit of one web-scraped ML training dataset reveals persistent PII after sanitization, which the authors combine with legal analysis to highlight privacy risks and advocate redefining 'publicly available' data for AI training.","context_count":1,"top_context_role":"background","top_context_polarity":"background","context_text":"can flow between actors and is aggregated by the dataset curator and then dispersed to various dataset users and model users. A detailed diagram is displayed in Figure 11. 4 RELATED WORK In this section, we highlight prior work both in the computer science and legal disciplines. 4.1 Data collection We draw upon many prior audits of datasets to inform our approach. Birhane et al. [16]provide a comprehensive overview of the AI audit ecosystem and analyze the priorities of existing data audits. On the web-scraping side, Dodge et al. [39] use keyword techniques and URL analysis to understand the text and websites within Common Crawl. Recent work has also audited the license and website terms of service restrictions of web-scraped machine learning datasets"},{"citing_arxiv_id":"2406.04952","ref_index":5,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Quantifying Geospatial in the Common Crawl Corpus","primary_cat":"cs.CL","submitted_at":"2024-06-07T14:16:37+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"Analysis estimates 18.7% of Common Crawl documents contain geospatial information like coordinates and addresses, with little difference by language.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2310.12508","ref_index":135,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"SalUn: Empowering Machine Unlearning via Gradient-based Weight Saliency in Both Image Classification and Generation","primary_cat":"cs.LG","submitted_at":"2023-10-19T06:17:17+00:00","verdict":"CONDITIONAL","verdict_confidence":"MODERATE","novelty_score":6.0,"formal_verification":"none","one_line_summary":"SalUn uses gradient-based weight saliency to achieve effective machine unlearning of data, classes, or concepts in image classification and generation, narrowing the gap to exact retraining.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2211.11018","ref_index":4,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"MagicVideo: Efficient Video Generation With Latent Diffusion Models","primary_cat":"cs.CV","submitted_at":"2022-11-20T16:40:31+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"MagicVideo generates 256x256 text-conditioned video clips via latent diffusion with a custom 3D U-Net, achieving roughly 64 times lower compute than prior video diffusion models.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2211.05100","ref_index":206,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"BLOOM: A 176B-Parameter Open-Access Multilingual Language Model","primary_cat":"cs.CL","submitted_at":"2022-11-09T18:48:09+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"BLOOM is a 176B-parameter open-access multilingual language model trained on the ROOTS corpus that achieves competitive performance on benchmarks, with improved results after multitask prompted finetuning.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2210.02303","ref_index":3,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Imagen Video: High Definition Video Generation with Diffusion Models","primary_cat":"cs.CV","submitted_at":"2022-10-05T14:41:38+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"Imagen Video generates high-definition text-conditional videos via a cascade of base and super-resolution diffusion models, achieving high fidelity and controllability.","context_count":1,"top_context_role":"method","top_context_polarity":"use_method","context_text":"lower and upper bounds on reverse process entropy (Sohl-Dickstein et al., 2015; Ho et al., 2020; Nichol & Dhariwal, 2021). This sampler can be formulated by using a reversed description of the forward process asq(zs|zt, x) =N (zs; ˜µs|t(zt, x), ˜σ2 s|tI) (notings