{"total":22,"items":[{"citing_arxiv_id":"2605.19042","ref_index":23,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Interference-Aware Multi-Task Unlearning","primary_cat":"cs.AI","submitted_at":"2026-05-18T19:05:40+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"Introduces interference-aware multi-task unlearning with task-aware gradient projection and instance-level gradient orthogonalization, reducing interference scores by 30.3% and 52.9% on vision benchmarks.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.18879","ref_index":5,"ref_count":2,"confidence":0.9,"is_internal_anchor":false,"paper_title":"ZeroUnlearn: Few-Shot Knowledge Unlearning in Large Language Models","primary_cat":"cs.LG","submitted_at":"2026-05-16T03:10:36+00:00","verdict":null,"verdict_confidence":null,"novelty_score":null,"formal_verification":null,"one_line_summary":null,"context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.07419","ref_index":17,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Incentivizing User Data Contributions for LLM Improvement under Withdrawal Rights","primary_cat":"cs.GT","submitted_at":"2026-05-08T08:14:29+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"Withdrawal rights paired with centralized cost-based assignment prevent subsidy waste by collecting data only when the improvement threshold is sustainably reachable, turning infeasible cases into null outcomes.","context_count":1,"top_context_role":"background","top_context_polarity":"background","context_text":"Dynamic voluntary contribution to a public project.The Review of Economic Studies, 67(2):327-358, 2000. [15] Todd Sandler. Collective action: fifty years later.Public Choice, 164:195-216, 2015. [16] Antonio A. Ginart, Melody Y . Guan, Gregory Valiant, and James Zou. Making ai forget you: Data deletion in machine learning.Advances in Neural Information Processing Systems, 32:3518-3531, 2019. [17] C. Guo, T. Goldstein, A. Hannun, and et al. Certified data removal from machine learning models.arXiv preprint arXiv:1911.03030, 2019. 10 [18] Dirk Bergemann, Alessandro Bonatti, and Tan Gan. The economics of social data.The RAND Journal of Economics, 53(2):263-296, 2022. [19] Jun Zhang, Yuxin Bi, Meng Cheng, and Qiang Yang. A survey on data markets."},{"citing_arxiv_id":"2605.05076","ref_index":34,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"High-Dimensional Statistics: Reflections on Progress and Open Problems","primary_cat":"math.ST","submitted_at":"2026-05-06T16:11:09+00:00","verdict":null,"verdict_confidence":null,"novelty_score":null,"formal_verification":null,"one_line_summary":null,"context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.03547","ref_index":41,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"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":"2604.23046","ref_index":8,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Shape of Memory: a Geometric Analysis of Machine Unlearning in Second-Order Optimizers","primary_cat":"cs.LG","submitted_at":"2026-04-24T22:10:01+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"Second-order optimizers retain residual geometric memory in their state after unlearning that first-order metrics miss, and only controlled eigendecay perturbations fully erase it.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2604.17396","ref_index":156,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Representation-Guided Parameter-Efficient LLM Unlearning","primary_cat":"cs.CL","submitted_at":"2026-04-19T11:59:58+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"REGLU guides LoRA-based unlearning via representation subspaces and orthogonal regularization to outperform prior methods on forget-retain trade-off in LLM benchmarks.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2604.13438","ref_index":9,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"WIN-U: Woodbury-Informed Newton-Unlearning as a retain-free Machine Unlearning Framework","primary_cat":"cs.LG","submitted_at":"2026-04-15T03:32:41+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"WIN-U delivers a retain-free unlearning update that approximates the gold-standard retrained model via a Woodbury-informed Newton step using only forget-set curvature information.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2604.12548","ref_index":35,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"DeepSeek Robustness Against Semantic-Character Dual-Space Mutated Prompt Injection","primary_cat":"cs.CR","submitted_at":"2026-04-14T10:20:15+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":4.0,"formal_verification":"none","one_line_summary":"Dual-space semantic-character mutations on prompts achieve higher misuse success rates against DeepSeek than single-space attacks alone.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2604.12348","ref_index":13,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"PrivEraserVerify: Efficient, Private, and Verifiable Federated Unlearning","primary_cat":"cs.LG","submitted_at":"2026-04-14T06:33:58+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"PrivEraserVerify unifies efficiency via adaptive checkpointing, privacy via layer-adaptive DP, and verifiability via fingerprints in federated unlearning, claiming 2-3x faster performance than retraining with formal guarantees.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2604.07386","ref_index":7,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Label Leakage Attacks in Machine Unlearning: A Parameter and Inversion-Based Approach","primary_cat":"cs.CR","submitted_at":"2026-04-08T03:57:48+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Parameter-difference and model-inversion attacks can identify forgotten classes after machine unlearning on standard image datasets.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2604.04800","ref_index":16,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Forgetting to Witness: Efficient Federated Unlearning and Its Visible Evaluation","primary_cat":"cs.LG","submitted_at":"2026-04-06T16:06:07+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"A complete pipeline for federated unlearning via knowledge distillation for efficient removal and a GAN-integrated classifier for visual evaluation of forgetting capacity.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2604.04030","ref_index":9,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"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":"baseline","top_context_polarity":"baseline","context_text":"The predictions made by the global model could potentially leak client information [7][8]. Therefore, a compelling need arises for a method to effectively eliminate a client's contribution from the trained global model. Two approaches exist to ensure that a global model forgets the contributions of a specific client. The first involves retraining the model from scratch after excluding 2 the data of the target user [9]. The second approach directly removes the user infor- mation from the trained model's parameters while preserving its overall utility[10]. However, retraining becomes impractical when dealing with large datasets and com- plex models due to its significant time and energy demands, high computational costs, and scalability challenges. Consequently, there is a growing interest in developing"},{"citing_arxiv_id":"2601.06163","ref_index":4,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"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":"2511.19339","ref_index":16,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"POUR: A Provably Optimal Method for Unlearning Representations via Neural Collapse","primary_cat":"cs.CV","submitted_at":"2025-11-24T17:38:53+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"POUR derives a provably optimal forgetting operator by showing that orthogonal projections of simplex equiangular tight frames remain ETFs in lower dimensions, enabling representation-level unlearning with closed-form and distillation variants.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2505.10859","ref_index":13,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"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},{"citing_arxiv_id":"2411.08443","ref_index":6,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Machine Unlearning on Pre-trained Models by Residual Feature Alignment Using LoRA","primary_cat":"cs.LG","submitted_at":"2024-11-13T08:56:35+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"A LoRA-based residual feature alignment method for efficient machine unlearning on pre-trained models by targeting zero residuals on retained data and shifted residuals on unlearned data.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2411.02622","ref_index":6,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"AdaProb: Efficient Machine Unlearning via Adaptive Probability","primary_cat":"cs.LG","submitted_at":"2024-11-04T21:27:06+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"AdaProb performs machine unlearning by substituting final-layer output probabilities with optimized uniform pseudo-probabilities and updating model weights.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2404.05868","ref_index":9,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Negative Preference Optimization: From Catastrophic Collapse to Effective Unlearning","primary_cat":"cs.LG","submitted_at":"2024-04-08T21:05:42+00:00","verdict":"CONDITIONAL","verdict_confidence":"LOW","novelty_score":8.0,"formal_verification":"none","one_line_summary":"NPO enables stable unlearning of 50%+ training data in LLMs on TOFU by making collapse exponentially slower than gradient ascent, preserving sensible outputs where prior methods fail.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2401.06121","ref_index":10,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"TOFU: A Task of Fictitious Unlearning for LLMs","primary_cat":"cs.LG","submitted_at":"2024-01-11T18:57:12+00:00","verdict":"CONDITIONAL","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"TOFU is a new benchmark with synthetic profiles and metrics demonstrating that existing unlearning algorithms for LLMs fail to achieve effective forgetting of targeted information.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2310.12508","ref_index":214,"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":"2308.05374","ref_index":203,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Trustworthy LLMs: a Survey and Guideline for Evaluating Large Language Models' Alignment","primary_cat":"cs.AI","submitted_at":"2023-08-10T06:43:44+00:00","verdict":"ACCEPT","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"Survey organizes LLM trustworthiness into seven categories and 29 sub-categories, measures eight sub-categories on popular models, and finds that more aligned models generally score higher but with varying effectiveness.","context_count":1,"top_context_role":"background","top_context_polarity":"background","context_text":"attacker can extract personal or sensitive information or private training samples from LLM's training data by querying LLMs alone. Researchers have proposed attacks that leverage the memorization effect of LLMs, usually growing with training sample repetition [193, 194]. Commonly used privacy-enhancing technologies (PETs) that defend against privacy attacks include differentially private training mechanisms [195, 196, 197, 198, 199], machine unlearning [200, 201, 202, 203, 204, 205], federated learning [206, 207, 208, 209, 210], and secure multi-party computation protocols [211, 212, 213, 214, 215, 216, 217, 218]. Note that although each of those privacy-enhancing techniques has a rich literature, the effectiveness and efficiency of them when applied to LLMs at a large scale is still unclear. 6 Fairness Due to the nature of training on crowdsourced and uncurated text corpora, it has been observed that LLMs can favor"}],"limit":50,"offset":0}