{"total":96,"rows":[{"id":"2408.15221","title":"LLM Defenses Are Not Robust to Multi-Turn Human Jailbreaks Yet","created":"2024-08-27","cited_by_count":151,"area":"safety"},{"id":"2410.07471","title":"SEAL: Safety-enhanced Aligned LLM Fine-tuning via Bilevel Data Selection","created":"2024-10-09","cited_by_count":64,"area":null},{"id":"2407.14937","title":"Operationalizing a Threat Model for Red-Teaming Large Language Models (LLMs)","created":"2024-07-20","cited_by_count":46,"area":"safety"},{"id":"2410.09760","title":"Targeted Vaccine: Safety Alignment for Large Language Models against Harmful Fine-Tuning via Layer-wise Perturbation","created":"2024-10-13","cited_by_count":37,"area":"alignment"},{"id":"2410.10014","title":"Safety-Aware Fine-Tuning of Large Language Models","created":"2024-10-13","cited_by_count":34,"area":null},{"id":"2504.01550","title":"Representation Bending for Large Language Model Safety","created":"2025-04-02","cited_by_count":30,"area":"safety"},{"id":"2406.11614","title":"Intrinsic Test of Unlearning Using Parametric Knowledge Traces","created":"2024-06-17","cited_by_count":29,"area":"safety"},{"id":"2502.01042","title":"SafeSwitch: Steering Unsafe LLM Behavior via Internal Activation Signals","created":"2025-02-03","cited_by_count":29,"area":"safety"},{"id":"2503.17239","title":"SafeMERGE: Preserving Safety Alignment in Fine-Tuned Large Language Models via Selective Layer-Wise Model Merging","created":"2025-03-21","cited_by_count":25,"area":"alignment"},{"id":"2507.18631","title":"Layer-Aware Representation Filtering: Purifying Finetuning Data to Preserve LLM Safety Alignment","created":"2025-07-24","cited_by_count":24,"area":"alignment"},{"id":"2410.10343","title":"Locking Down the Finetuned LLMs Safety","created":"2024-10-14","cited_by_count":22,"area":null},{"id":"2505.06843","title":"Benign Samples Matter! Fine-tuning On Outlier Benign Samples Severely Breaks Safety","created":"2025-05-11","cited_by_count":22,"area":"safety"},{"id":"2502.17601","title":"Representation Engineering for Large-Language Models: Survey and Research Challenges","created":"2025-02-24","cited_by_count":21,"area":"mechinterp"},{"id":"2506.05346","title":"Why LLM Safety Guardrails Collapse After Fine-tuning: A Similarity Analysis Between Alignment and Fine-tuning Datasets","created":"2025-06-05","cited_by_count":20,"area":"alignment"},{"id":"2506.08473","title":"AsFT: Anchoring Safety During LLM Fine-Tuning Within Narrow Safety Basin","created":"2025-06-10","cited_by_count":19,"area":null},{"id":"2505.12038","title":"Safe Delta: Consistently Preserving Safety when Fine-Tuning LLMs on Diverse Datasets","created":"2025-05-17","cited_by_count":18,"area":null},{"id":"2505.23559","title":"SafeScientist: Toward Risk-Aware Scientific Discoveries by LLM Agents","created":"2025-05-29","cited_by_count":16,"area":"agents"},{"id":"2507.21182","title":"SDD: Self-Degraded Defense against Malicious Fine-tuning","created":"2025-07-27","cited_by_count":13,"area":"alignment"},{"id":"2601.21571","title":"Shaping capabilities with token-level data filtering","created":"2026-01-29","cited_by_count":13,"area":null},{"id":"2506.15606","title":"LoX: Low-Rank Extrapolation Robustifies LLM Safety Against Fine-tuning","created":"2025-06-18","cited_by_count":12,"area":"safety"},{"id":"2505.16559","title":"CTRAP: Embedding Collapse Trap to Safeguard Large Language Models from Harmful Fine-Tuning","created":"2025-05-22","cited_by_count":11,"area":"safety"},{"id":"2506.06278","title":"Distillation Robustifies Unlearning","created":"2025-06-06","cited_by_count":11,"area":"safety"},{"id":"2508.17158","title":"Towards Safeguarding LLM Fine-tuning APIs against Cipher Attacks","created":"2025-08-23","cited_by_count":11,"area":null},{"id":"2502.14828","title":"Fundamental Limitations in Pointwise Defences of LLM Finetuning APIs","created":"2025-02-20","cited_by_count":10,"area":null},{"id":"2502.15435","title":"Single-pass Detection of Jailbreaking Input in Large Language Models","created":"2025-02-21","cited_by_count":10,"area":"alignment"},{"id":"2505.12186","title":"Self-Destructive Language Model","created":"2025-05-18","cited_by_count":10,"area":"safety"},{"id":"2507.11630","title":"Jailbreak-Tuning: Models Efficiently Learn Jailbreak Susceptibility","created":"2025-07-15","cited_by_count":9,"area":"safety"},{"id":"2510.14301","title":"A Guardrail for Safety Preservation: When Safety-Sensitive Subspace Meets Harmful-Resistant Null-Space","created":"2025-10-16","cited_by_count":9,"area":"alignment"},{"id":"2506.06414","title":"Benchmarking Misuse Mitigation Against Covert Adversaries","created":"2025-06-06","cited_by_count":8,"area":null},{"id":"2508.20697","title":"Token Buncher: Shielding LLMs from Harmful Reinforcement Learning Fine-Tuning","created":"2025-08-28","cited_by_count":8,"area":null},{"id":"2509.22745","title":"Defending MoE LLMs against Harmful Fine-Tuning via Safety Routing Alignment","created":"2025-09-26","cited_by_count":8,"area":"moe"},{"id":"2510.07835","title":"MetaDefense: Defending Finetuning-based Jailbreak Attack Before and During Generation","created":"2025-10-09","cited_by_count":8,"area":"safety"},{"id":"2606.23075","title":"Safety in Self-Evolving LLM Agent Systems: Threats, Amplification, and Case Studies","created":"2026-06-22","cited_by_count":8,"area":"agents"},{"id":"2601.10141","title":"Understanding and Preserving Safety in Fine-Tuned LLMs","created":"2026-01-15","cited_by_count":7,"area":"safety"},{"id":"2604.01476","title":"From Rebound to Remedy: Understanding and Mitigating Reward Hacking via Representation Engineering","created":"2026-04-01","cited_by_count":7,"area":"alignment"},{"id":"2502.13946","title":"Why Safeguarded Ships Run Aground? Aligned Large Language Models' Safety Mechanisms Tend to Be Anchored in The Template Region","created":"2025-02-19","cited_by_count":6,"area":"alignment"},{"id":"2505.14185","title":"Safety Subspaces are Not Linearly Distinct: A Fine-Tuning Case Study","created":"2025-05-20","cited_by_count":6,"area":"alignment"},{"id":"2506.07452","title":"When Style Breaks Safety: Defending LLMs Against Superficial Style Alignment","created":"2025-06-09","cited_by_count":6,"area":"safety"},{"id":"2603.00498","title":"Antibody: Strengthening Defense Against Harmful Fine-Tuning for Large Language Models via Attenuating Harmful Gradient Influence","created":"2026-02-28","cited_by_count":6,"area":"safety"},{"id":"2506.10236","title":"Prompt Attacks Reveal Superficial Knowledge Removal in Unlearning Methods","created":"2025-06-11","cited_by_count":5,"area":"safety"},{"id":"2504.11364","title":"Offline Learning and Forgetting for Reasoning with Large Language Models","created":"2025-04-15","cited_by_count":4,"area":"lr"},{"id":"2504.14810","title":"DONOD: Efficient and Generalizable Instruction Fine-Tuning for LLMs via Model-Intrinsic Dataset Pruning","created":"2025-04-21","cited_by_count":4,"area":null},{"id":"2601.22313","title":"Hair-Trigger Alignment: Black-Box Evaluation Cannot Guarantee Post-Update Alignment","created":"2026-01-29","cited_by_count":4,"area":null},{"id":"2602.05228","title":"Surgery: Mitigating Harmful Fine-Tuning for Large Language Models via Attention Sink","created":"2026-02-05","cited_by_count":4,"area":"alignment"},{"id":"2602.10568","title":"Gauss-Newton Unlearning for the LLM Era","created":"2026-02-11","cited_by_count":4,"area":"safety"},{"id":"2505.04578","title":"Fight Fire with Fire: Defending Against Malicious RL Fine-Tuning via Reward Neutralization","created":"2025-05-07","cited_by_count":3,"area":"safety"},{"id":"2510.15499","title":"HarmRLVR: Weaponizing Verifiable Rewards for Harmful LLM Alignment","created":"2025-10-17","cited_by_count":3,"area":"alignment"},{"id":"2512.01899","title":"Provably Safe Model Updates","created":"2025-12-01","cited_by_count":3,"area":null},{"id":"2602.08813","title":"Robust Policy Optimization to Prevent Catastrophic Forgetting","created":"2026-02-09","cited_by_count":3,"area":"alignment"},{"id":"2502.05242","title":"Beyond External Monitors: Enhancing Transparency of Large Language Models for Easier Monitoring","created":"2025-02-07","cited_by_count":2,"area":null},{"id":"2506.07356","title":"Safety-Aligned Weights Are Not Enough: Refusal-Teacher-Guided Finetuning Enhances Safety and Downstream Performance under Harmful Finetuning Attacks","created":"2025-06-09","cited_by_count":2,"area":"safety"},{"id":"2507.11544","title":"The Safety Gap Toolkit: Evaluating Hidden Dangers of Open-Source Models","created":"2025-07-08","cited_by_count":2,"area":null},{"id":"2510.01342","title":"Fine-Tuning Jailbreaks under Highly Constrained Black-Box Settings: A Three-Pronged Approach","created":"2025-10-01","cited_by_count":2,"area":"safety"},{"id":"2510.07775","title":"The Unintended Trade-off of AI Alignment:Balancing Hallucination Mitigation and Safety in LLMs","created":"2025-10-09","cited_by_count":2,"area":"safety"},{"id":"2601.07200","title":"Safeguarding LLM Fine-tuning via Push-Pull Distributional Alignment","created":"2026-01-12","cited_by_count":2,"area":"alignment"},{"id":"2605.14605","title":"A Few Steps Further: Why Defenses Against Malicious Finetuning Erode Under Continued Training","created":"2026-05-14","cited_by_count":2,"area":null},{"id":"2606.07688","title":"TRACER: Token ReAssignment for Concept ERasure in Generative Recommendation","created":"2026-06-05","cited_by_count":2,"area":"safety"},{"id":"2606.08044","title":"When Behavioral Safety Evaluation Fails: A Representation-Level Perspective","created":"2026-06-06","cited_by_count":2,"area":"safety"},{"id":"2410.04155","title":"Toxic Subword Pruning for Dialogue Response Generation on Large Language Models","created":"2024-10-05","cited_by_count":1,"area":null},{"id":"2502.01472","title":"FALCON: Fine-grained Activation Manipulation by Contrastive Orthogonal Unalignment for Large Language Model","created":"2025-02-03","cited_by_count":1,"area":"alignment"},{"id":"2505.09500","title":"Layered Unlearning for Adversarial Relearning","created":"2025-05-14","cited_by_count":1,"area":"safety"},{"id":"2507.13598","title":"GIFT: Gradient-aware Immunization of diffusion models against malicious Fine-Tuning with safe concepts retention","created":"2025-07-18","cited_by_count":1,"area":"diffusion"},{"id":"2512.04144","title":"RippleBench: Capturing Ripple Effects Using Existing Knowledge Repositories","created":"2025-12-03","cited_by_count":1,"area":"safety"},{"id":"2601.21702","title":"Beyond Forgetting: Representation Misdirection Elicits Controllable Side Behaviors and Capabilities","created":"2026-01-29","cited_by_count":1,"area":"safety"},{"id":"2603.04355","title":"Efficient Refusal Ablation in LLM through Optimal Transport","created":"2026-03-04","cited_by_count":1,"area":"alignment"},{"id":"2603.24857","title":"AI Security in the Foundation Model Era: A Comprehensive Survey from a Unified Perspective","created":"2026-03-25","cited_by_count":1,"area":"safety"},{"id":"2604.09688","title":"Immunizing 3D Gaussian Generative Models Against Unauthorized Fine-Tuning via Attribute-Space Traps","created":"2026-04-06","cited_by_count":1,"area":null},{"id":"2605.11685","title":"Robust LLM Unlearning Against Relearning Attacks: The Minor Components in Representations Matter","created":"2026-05-12","cited_by_count":1,"area":"safety"},{"id":"2605.25073","title":"Security in the Fine-Tuning Lifecycle of Large Language Models: Threats, Defenses,Evaluation, and Future Directions","created":"2026-05-24","cited_by_count":1,"area":"safety"},{"id":"2605.26526","title":"Open-Weight LLM Fine-Tuning Defenses are Susceptible to Simple Attacks","created":"2026-05-26","cited_by_count":1,"area":"safety"},{"id":"2605.29396","title":"Aligned but Fragile: Enhancing LLM Safety Robustness via Zeroth-Order Optimization","created":"2026-05-28","cited_by_count":1,"area":"alignment"},{"id":"2606.26036","title":"Detect, Unlearn, Restore: Defending Text Summarization Models Against Data Poisoning","created":"2026-06-24","cited_by_count":1,"area":"safety"},{"id":"2602.03906","title":"GeoIB: Geometry-Aware Information Bottleneck via Statistical-Manifold Compression","created":"2026-02-03","cited_by_count":0,"area":null},{"id":"2602.17546","title":"Learning to Stay Safe: Adaptive Regularization Against Safety Degradation during Fine-Tuning","created":"2026-02-19","cited_by_count":0,"area":null},{"id":"2603.04407","title":"Semantic Containment as a Fundamental Property of Emergent Misalignment","created":"2026-02-02","cited_by_count":0,"area":null},{"id":"2604.00324","title":"The Persistent Vulnerability of Aligned AI Systems","created":"2026-03-31","cited_by_count":0,"area":null},{"id":"2604.12469","title":"Analyzing the Effect of Noise in LLM Fine-tuning","created":"2026-04-14","cited_by_count":0,"area":null},{"id":"2604.17691","title":"SafeAnchor: Preventing Cumulative Safety Erosion in Continual Domain Adaptation of Large Language Models","created":"2026-04-20","cited_by_count":0,"area":"alignment"},{"id":"2605.05995","title":"Safety Anchor: Defending Harmful Fine-tuning via Geometric Bottlenecks","created":"2026-05-07","cited_by_count":0,"area":"alignment"},{"id":"2605.10575","title":"Acceptance Cards:A Four-Diagnostic Standard for Safe Fine-Tuning Defense Claims","created":"2026-05-11","cited_by_count":0,"area":null},{"id":"2605.10777","title":"Locking Pretrained Weights via Deep Low-Rank Residual Distillation","created":"2026-05-11","cited_by_count":0,"area":null},{"id":"2605.14194","title":"GradShield: Alignment Preserving Finetuning","created":"2026-05-13","cited_by_count":0,"area":null},{"id":"2605.24550","title":"Jailbreak to Protect: Buffering and Reinforcing via Temporary Jailbreaking for Safe Fine-Tuning in Large Language Models","created":"2026-05-23","cited_by_count":0,"area":"safety"},{"id":"2606.07970","title":"Defending Against Malicious Finetuning by Scaling Train-time Adversarial Attacks","created":"2026-06-06","cited_by_count":0,"area":"alignment"},{"id":"2606.09559","title":"Safe-RULE: Safe Reinforcement UnLEarning","created":"2026-06-08","cited_by_count":0,"area":"safety"},{"id":"2606.28525","title":"A Gravitational Interpretation of Safety Reversion under Fine-Tuning","created":"2026-06-26","cited_by_count":0,"area":"safety"},{"id":"2607.11475","title":"HyperSafe: Inference-Time Safety Recovery for Fine-Tuned Language Models","created":"2026-07-13","cited_by_count":0,"area":"alignment"},{"id":"2608.23497","title":"Mitigating Reasoning-Induced Misalignment via Safety-Direction Penalty","created":"2026-08-24","cited_by_count":0,"area":"lr"},{"id":"2609.05794","title":"Bait-and-Recover: Poisoning Internal Refusal Signals to Defend LLMs against White-Box Editing Jailbreaks","created":"2026-09-05","cited_by_count":0,"area":"safety"},{"id":"2609.06934","title":"The Geometry of Refusal: Why Post-Hoc Safety Is Fragile and Pretraining-Time Safety Persists","created":"2026-09-07","cited_by_count":0,"area":"alignment"},{"id":"2609.22153","title":"SafeTune: A Unified Faithful Library for Auditing and Repairing Safety Drift in Fine-Tuned LLMs","created":"2026-08-26","cited_by_count":0,"area":null},{"id":"2609.29960","title":"Beyond Average Safety: Chance-Constrained LLM Fine-tuning","created":"2026-09-24","cited_by_count":0,"area":null},{"id":"2609.39866","title":"Preemptive LLM Unlearning against Forbidden Capability Acquisition via Gradient Sealing","created":"2026-09-30","cited_by_count":0,"area":"safety"},{"id":"2610.00320","title":"Refusal Localizes, the Damage Relocates: Safety Layers Under Few-Sample Fine-Tuning","created":"2026-09-29","cited_by_count":0,"area":"alignment"},{"id":"2610.04316","title":"From Latent Space to Jacobian Space: Measuring, Evading, and Training Against Safety-Content Accessibility","created":"2026-10-03","cited_by_count":0,"area":null},{"id":"2610.10345","title":"SLDR: Defending Against Malicious Fine-tuning via Selective Layers Recovery and Dynamic Routing","created":"2026-10-07","cited_by_count":0,"area":"alignment"}],"next_cursor":null}