{"total":24,"items":[{"citing_arxiv_id":"2605.22311","ref_index":9,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"PIU: Proximity-guided Identity Unlearning in ID-Conditioned Diffusion Models","primary_cat":"cs.CV","submitted_at":"2026-05-21T10:55:56+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"PIU suppresses target identity generation in Arc2Face by replacing it with a proximity-selected anchor identity through localized fine-tuning of cross-attention layers while preserving output quality for other identities.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.19750","ref_index":8,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"CPC-VAR:Continual Personalized and Compositional Generation in Visual Autoregressive Models","primary_cat":"cs.CV","submitted_at":"2026-05-19T12:18:15+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"CPC-VAR adds Gradient-based Concept Neuron Selection for continual single-concept learning and a context-aware multi-branch composition strategy to reduce forgetting and entanglement in VAR-based personalized image generation.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.18253","ref_index":23,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Machine Unlearning for Masked Diffusion Language Models","primary_cat":"cs.CL","submitted_at":"2026-05-18T11:54:11+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"MDU minimizes forward KL divergence from prompt-conditional to prompt-masked unconditional predictions at masked positions to unlearn knowledge in MDLMs while trading off privacy and utility via temperature scaling.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.15737","ref_index":11,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"BARRIER: Bounded Activation Regions for Robust Information Erasure","primary_cat":"cs.CV","submitted_at":"2026-05-15T08:46:02+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"BARRIER applies interval arithmetic to SVD-based activation projections to create bounded forget regions that enable aggressive unlearning while providing formal protection for retain distributions via tail bounds on functional drift.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.11170","ref_index":40,"ref_count":1,"confidence":0.9,"is_internal_anchor":true,"paper_title":"Unlearning with Asymmetric Sources: Improved Unlearning-Utility Trade-off with Public Data","primary_cat":"cs.LG","submitted_at":"2026-05-11T19:28:33+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"Asymmetric Langevin Unlearning uses public data to suppress unlearning noise costs by O(1/n_pub²), enabling practical mass unlearning with preserved utility under distribution mismatch.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.08730","ref_index":22,"ref_count":1,"confidence":0.9,"is_internal_anchor":true,"paper_title":"Classification-Head Bias in Class-Level Machine Unlearning: Diagnosis, Mitigation, and Evaluation","primary_cat":"cs.LG","submitted_at":"2026-05-09T06:33:42+00:00","verdict":"CONDITIONAL","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"Class-level unlearning shortcuts via bias suppression in the classification head; new bias-aware training mechanisms and bias-specific metrics are introduced to diagnose and reduce this dependence.","context_count":1,"top_context_role":"background","top_context_polarity":"background","context_text":"sign incorrect or random labels to forget samples, forcing the model to fit noisy supervision and thereby weakening its predictive ability on the forgotten classes [17]. Saliency- based unlearning methods, such as SalUn, identify param- eters that are more relevant to the forget set and selec- tively update them to improve unlearning efficiency and stability [22]. Distillation-based methods, such as SCRUB, formulate unlearning as a selective knowledge distillation problem, encouraging the unlearned model to deviate from the teacher model on the forget set while preserving similar behavior on the retain set [23]. Noise-based methods, such as UNSIR, first generate error-maximizing noise to impair the model components associated with forgotten classes and"},{"citing_arxiv_id":"2605.05909","ref_index":38,"ref_count":1,"confidence":0.9,"is_internal_anchor":true,"paper_title":"Null Space Constrained Contrastive Visual Forgetting for MLLM Unlearning","primary_cat":"cs.AI","submitted_at":"2026-05-07T09:18:58+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"A contrastive visual forgetting technique constrained to the null space of retained knowledge enables targeted unlearning of visual concepts in MLLMs while preserving non-target visual and all textual knowledge.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.03547","ref_index":50,"ref_count":1,"confidence":0.9,"is_internal_anchor":true,"paper_title":"Erase Persona, Forget Lore: Benchmarking Multimodal Copyright Unlearning in Large Vision Language Models","primary_cat":"cs.CV","submitted_at":"2026-05-05T09:18:22+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"CoVUBench is the first benchmark framework for evaluating multimodal copyright unlearning in LVLMs via synthetic data, systematic variations, and a dual protocol for forgetting efficacy and utility preservation.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2604.27804","ref_index":15,"ref_count":1,"confidence":0.9,"is_internal_anchor":true,"paper_title":"Machine Unlearning for Class Removal through SISA-based Deep Neural Network Architectures","primary_cat":"cs.CV","submitted_at":"2026-04-30T12:47:50+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"A modified SISA architecture with replay and gating achieves effective class removal from trained CNNs on image datasets while preserving accuracy and cutting retraining costs.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2604.25119","ref_index":8,"ref_count":1,"confidence":0.9,"is_internal_anchor":true,"paper_title":"Evaluation without Generation: Non-Generative Assessment of Harmful Model Specialization with Applications to CSAM","primary_cat":"cs.LG","submitted_at":"2026-04-28T01:54:25+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Gaussian probing infers harmful model specialization from parameter perturbations and internal representation responses to Gaussian latent ensembles rather than from generated outputs.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2604.24022","ref_index":14,"ref_count":1,"confidence":0.9,"is_internal_anchor":true,"paper_title":"IPRU: Input-Perturbation-based Radio Frequency Fingerprinting Unlearning for LAWNs","primary_cat":"eess.SP","submitted_at":"2026-04-27T04:10:31+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"IPRU erases target AAV radio fingerprints via an optimized input perturbation vector, delivering 1.41% unlearning accuracy, 99.41% remaining accuracy, full membership-inference resistance, and 5.79X speedup over retraining.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2604.15829","ref_index":7,"ref_count":1,"confidence":0.9,"is_internal_anchor":true,"paper_title":"Beyond Text Prompts: Precise Concept Erasure through Text-Image Collaboration","primary_cat":"cs.CV","submitted_at":"2026-04-17T08:32:39+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"TICoE achieves more precise and faithful concept erasure in text-to-image models by collaborating text and image data through a convex manifold and hierarchical learning, outperforming prior methods.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2604.15166","ref_index":13,"ref_count":2,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Class Unlearning via Depth-Aware Removal of Forget-Specific Directions","primary_cat":"cs.CV","submitted_at":"2026-04-16T15:46:02+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"DAMP performs one-shot class unlearning by depth-aware projection removal of forget-specific directions, producing forgetting behavior closer to retraining from scratch than prior methods on image classification tasks.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2604.12686","ref_index":13,"ref_count":1,"confidence":0.9,"is_internal_anchor":true,"paper_title":"BID-LoRA: A Parameter-Efficient Framework for Continual Learning and Unlearning","primary_cat":"cs.LG","submitted_at":"2026-04-14T12:57:56+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"BID-LoRA uses bi-directional low-rank adapters with retain/new/unlearn pathways and escape unlearning to enable continual learning and unlearning while minimizing knowledge leakage and parameter updates.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2604.09405","ref_index":12,"ref_count":1,"confidence":0.9,"is_internal_anchor":true,"paper_title":"EGLOCE: Training-Free Energy-Guided Latent Optimization for Concept Erasure","primary_cat":"cs.CV","submitted_at":"2026-04-10T15:19:02+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"EGLOCE erases target concepts in diffusion models at inference time by optimizing latents with dual energy guidance that repels unwanted concepts while retaining prompt alignment.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2604.09391","ref_index":13,"ref_count":1,"confidence":0.9,"is_internal_anchor":true,"paper_title":"Efficient Unlearning through Maximizing Relearning Convergence Delay","primary_cat":"cs.LG","submitted_at":"2026-04-10T15:06:50+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"The Influence Eliminating Unlearning framework maximizes relearning convergence delay via weight decay and noise injection to remove the influence of a forgetting set while preserving accuracy on retained data.","context_count":1,"top_context_role":"baseline","top_context_polarity":"baseline","context_text":"should closely match the retraining model in behavior, re- flected in comparable accuracy across the retaining, forget- ting, and testing datasets. In image generation tasks, the FID measures how closely the distribution of generated images aligns with that of real images, and it is the current standard for evaluating the qual- ity of generative models [13, 28, 53]. A lower FID score indicates more realistically generated images, reflecting the effectiveness of the generative model. In the context of un- learning NSFW (not safe for work) content, previous works [13, 53] evaluate a model's ability to generate harmful con- tent by employing detection models that assess the level of harmfulness in the generated images."},{"citing_arxiv_id":"2604.08111","ref_index":10,"ref_count":1,"confidence":0.9,"is_internal_anchor":true,"paper_title":"Bias Redistribution in Visual Machine Unlearning: Does Forgetting One Group Harm Another?","primary_cat":"cs.LG","submitted_at":"2026-04-09T11:29:36+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Unlearning a demographic group in CLIP models redistributes bias primarily along gender boundaries rather than eliminating it.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2604.07962","ref_index":10,"ref_count":1,"confidence":0.9,"is_internal_anchor":true,"paper_title":"Is your algorithm unlearning or untraining?","primary_cat":"cs.LG","submitted_at":"2026-04-09T08:24:52+00:00","verdict":"CONDITIONAL","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"Machine unlearning conflates reversing the influence of specific training examples (untraining) with removing the full underlying distribution or behavior (unlearning).","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2604.04575","ref_index":5,"ref_count":1,"confidence":0.9,"is_internal_anchor":true,"paper_title":"Erasure or Erosion? Evaluating Compositional Degradation in Unlearned Text-To-Image Diffusion Models","primary_cat":"cs.CV","submitted_at":"2026-04-06T10:16:42+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Unlearning methods that strongly erase concepts from text-to-image diffusion models consistently degrade performance on attribute binding, spatial reasoning, and counting tasks.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2604.04030","ref_index":24,"ref_count":1,"confidence":0.9,"is_internal_anchor":true,"paper_title":"Jellyfish: Zero-Shot Federated Unlearning Scheme with Knowledge Disentanglement","primary_cat":"cs.CR","submitted_at":"2026-04-05T09:13:12+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Jellyfish enables zero-shot federated unlearning through synthetic proxy data generation, channel-restricted knowledge disentanglement, and a composite loss with repair to forget target data while retaining model utility.","context_count":1,"top_context_role":"background","top_context_polarity":"background","context_text":"machine learning. In: International Conference on Machine Learning, pp. 6028- 6073 (2023). PMLR [23] Chundawat, V.S., Tarun, A.K., Mandal, M., Kankanhalli, M.: Can bad teaching induce forgetting? unlearning in deep networks using an incompetent teacher. In: Proceedings of the AAAI Conference on Artificial Intelligence, vol. 37, pp. 7210-7217 (2023) [24] Fan, C., Liu, J., Zhang, Y., Wong, E., Wei, D., Liu, S.: Salun: Empowering machine unlearning via gradient-based weight saliency in both image classification and generation. arXiv preprint arXiv:2310.12508 (2023) [25] Graves, L., Nagisetty, V., Ganesh, V.: Amnesiac machine learning. In: Proceedings of the AAAI Conference on Artificial Intelligence, vol."},{"citing_arxiv_id":"2604.04982","ref_index":16,"ref_count":1,"confidence":0.9,"is_internal_anchor":true,"paper_title":"CURE:Circuit-Aware Unlearning for LLM-based Recommendation","primary_cat":"cs.IR","submitted_at":"2026-04-04T23:13:55+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"CURE disentangles LLM recommendation circuits into forget-specific, retain-specific, and task-shared modules with tailored update rules to achieve more effective unlearning than weighted baselines.","context_count":1,"top_context_role":"background","top_context_polarity":"background","context_text":"𝑣1→𝑣2 −𝑚𝑣1→𝑣2 )⊤ 𝜕Δ(𝑥 𝑢 ) 𝜕𝑣2 . (7) The second term on the right-hand side serves as an estimate of𝐼(𝑒) . However, unlike circuit detection in reasoning tasks, constructing the corrupt samplex ∗ 𝑢 in LLMRec is non-trivial. Two challenges arise: 1) The inputx 𝑢 can be substantially longer, as it summarizes a user's historical interactions. 2) Prior work [16, 27] requiresx 𝑢 and its corrupt counterpartx ∗ 𝑢 to differ minimally at the input level while inducing a significant change in prediction. As a result, it is difficult to pinpoint which tokens inx 𝑢 should be modified to constructx ∗ 𝑢. Formally,x ∗ 𝑢 should satisfy the following constraints x∗ 𝑢 =arg min x Δ(x)s.t.∥x−x 𝑢 ∥0 ≤𝐾 .(8) Here, 𝐾 controls the number of items to be replaced."},{"citing_arxiv_id":"2601.06163","ref_index":3,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Forget-It-All: Multi-Concept Machine Unlearning via Concept-Aware Neuron Masking","primary_cat":"cs.CV","submitted_at":"2026-01-07T00:13:36+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"FIA uses contrastive concept saliency and temporal-spatial neuron identification to build unified masks that erase multiple target concepts while preserving general generation quality in diffusion models.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2511.12968","ref_index":6,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"GrOCE:Graph-Guided Online Concept Erasure for Text-to-Image Diffusion Models","primary_cat":"cs.CV","submitted_at":"2025-11-17T04:47:16+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"GrOCE uses dynamic semantic graphs for online, training-free erasure of target concepts from diffusion model prompts via cluster identification and selective severing.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2505.10859","ref_index":6,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Exploring Nonlinear Pathway in Parameter Space for Machine Unlearning","primary_cat":"cs.AI","submitted_at":"2025-05-16T04:56:47+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"MCU applies mode connectivity to trace nonlinear unlearning pathways in parameter space, adds a parameter mask and adaptive penalty, and produces a range of unlearning models that plug into existing methods.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null}],"limit":50,"offset":0}