{"total":88,"rows":[{"id":"2501.17148","title":"AxBench: Steering LLMs? Even Simple Baselines Outperform Sparse Autoencoders","created":"2025-01-28","cited_by_count":218,"area":"mechinterp"},{"id":"2407.15549","title":"Latent Adversarial Training Improves Robustness to Persistent Harmful Behaviors in LLMs","created":"2024-07-22","cited_by_count":173,"area":"safety"},{"id":"2504.15585","title":"A Comprehensive Survey in LLM(-Agent) Full Stack Safety: Data, Training and Deployment","created":"2025-04-22","cited_by_count":116,"area":"lr"},{"id":"2405.15589","title":"Efficient Adversarial Training in LLMs with Continuous Attacks","created":"2024-05-24","cited_by_count":113,"area":"safety"},{"id":"2405.14577","title":"Representation Noising: A Defence Mechanism Against Harmful Finetuning","created":"2024-05-23","cited_by_count":96,"area":"safety"},{"id":"2409.01586","title":"Booster: Tackling Harmful Fine-tuning for Large Language Models via Attenuating Harmful Perturbation","created":"2024-09-03","cited_by_count":92,"area":null},{"id":"2409.18169","title":"Harmful Fine-tuning Attacks and Defenses for Large Language Models: A Survey","created":"2024-09-26","cited_by_count":90,"area":"alignment"},{"id":"2408.09600","title":"Antidote: Post-fine-tuning Safety Alignment for Large Language Models against Harmful Fine-tuning","created":"2024-08-18","cited_by_count":72,"area":"alignment"},{"id":"2409.20089","title":"Robust LLM safeguarding via refusal feature adversarial training","created":"2024-09-30","cited_by_count":64,"area":null},{"id":"2601.04603","title":"Constitutional Classifiers++: Efficient Production-Grade Defenses against Universal Jailbreaks","created":"2026-01-08","cited_by_count":41,"area":"safety"},{"id":"2502.05209","title":"Model Tampering Attacks Enable More Rigorous Evaluations of LLM Capabilities","created":"2025-02-03","cited_by_count":39,"area":null},{"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":"2410.02760","title":"Erasing Conceptual Knowledge from Language Models","created":"2024-10-03","cited_by_count":34,"area":"safety"},{"id":"2410.04332","title":"Gradient Routing: Masking Gradients to Localize Computation in Neural Networks","created":"2024-10-06","cited_by_count":34,"area":null},{"id":"2410.08811","title":"PoisonBench: Assessing Large Language Model Vulnerability to Data Poisoning","created":"2024-10-11","cited_by_count":33,"area":"safety"},{"id":"2409.07985","title":"Games for AI Control: Models of Safety Evaluations of AI Deployment Protocols","created":"2024-09-12","cited_by_count":31,"area":"safety"},{"id":"2407.15211","title":"Failures to Find Transferable Image Jailbreaks Between Vision-Language Models","created":"2024-07-21","cited_by_count":30,"area":"mllm"},{"id":"2408.10682","title":"Towards Robust Knowledge Unlearning: An Adversarial Framework for Assessing and Improving Unlearning Robustness in Large Language Models","created":"2024-08-20","cited_by_count":27,"area":"safety"},{"id":"2502.20968","title":"Beware of Your Po! Measuring and Mitigating AI Safety Risks in Role-Play Fine-Tuning of LLMs","created":"2025-02-28","cited_by_count":22,"area":null},{"id":"2501.18100","title":"Panacea: Mitigating Harmful Fine-tuning for Large Language Models via Post-fine-tuning Perturbation","created":"2025-01-30","cited_by_count":20,"area":null},{"id":"2603.06594","title":"A Coin Flip for Safety: LLM Judges Fail to Reliably Measure Adversarial Robustness","created":"2026-02-04","cited_by_count":20,"area":"safety"},{"id":"2406.11096","title":"The Potential and Challenges of Evaluating Attitudes, Opinions, and Values in Large Language Models","created":"2024-06-16","cited_by_count":19,"area":null},{"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":"2410.13211","title":"Estimating the Probabilities of Rare Outputs in Language Models","created":"2024-10-17","cited_by_count":16,"area":null},{"id":"2510.27062","title":"Consistency Training Helps Stop Sycophancy and Jailbreaks","created":"2025-10-31","cited_by_count":16,"area":"safety"},{"id":"2502.02260","title":"Position: Adversarial ML for LLMs Is Not Making Any Progress","created":"2025-02-04","cited_by_count":15,"area":null},{"id":"2502.04204","title":"Short-length Adversarial Training Helps LLMs Defend Long-length Jailbreak Attacks: Theoretical and Empirical Evidence","created":"2025-02-06","cited_by_count":14,"area":"safety"},{"id":"2506.24068","title":"STACK: Adversarial Attacks on LLM Safeguard Pipelines","created":"2025-06-30","cited_by_count":14,"area":null},{"id":"2505.24445","title":"Learning Safety Constraints for Large Language Models","created":"2025-05-30","cited_by_count":13,"area":"safety"},{"id":"2506.12707","title":"SecurityLingua: Efficient Defense of LLM Jailbreak Attacks via Security-Aware Prompt Compression","created":"2025-06-15","cited_by_count":13,"area":"inference"},{"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":"2505.16947","title":"MixAT: Combining Continuous and Discrete Adversarial Training for LLMs","created":"2025-05-22","cited_by_count":11,"area":"safety"},{"id":"2508.07172","title":"Gradient Surgery for Safe LLM Fine-Tuning","created":"2025-08-10","cited_by_count":11,"area":"alignment"},{"id":"2512.10150","title":"Unforgotten Safety: Preserving Safety Alignment of Large Language Models with Continual Learning","created":"2025-12-10","cited_by_count":10,"area":"alignment"},{"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":"2502.06470","title":"A Survey of Theory of Mind in Large Language Models: Evaluations, Representations, and Safety Risks","created":"2025-02-10","cited_by_count":8,"area":null},{"id":"2502.11910","title":"Adversarial Alignment for LLMs Requires Simpler, Reproducible, and More Measurable Objectives","created":"2025-02-17","cited_by_count":8,"area":null},{"id":"2506.11938","title":"Improving Large Language Model Safety with Contrastive Representation Learning","created":"2025-06-13","cited_by_count":8,"area":"mechinterp"},{"id":"2501.10639","title":"Latent-space adversarial training with post-aware calibration for defending large language models against jailbreak attacks","created":"2025-01-18","cited_by_count":7,"area":"alignment"},{"id":"2505.24369","title":"Adversarial Preference Learning for Robust LLM Alignment","created":"2025-05-30","cited_by_count":7,"area":"alignment"},{"id":"2510.21910","title":"Adversarial D\\'ej\\`a Vu: Jailbreak Dictionary Learning for Stronger Generalization to Unseen Attacks","created":"2025-10-24","cited_by_count":7,"area":"safety"},{"id":"2605.11161","title":"Interpretability Can Be Actionable","created":"2026-05-11","cited_by_count":7,"area":null},{"id":"2409.03274","title":"Recent Advances in Attack and Defense Approaches of Large Language Models","created":"2024-09-05","cited_by_count":6,"area":null},{"id":"2502.19537","title":"No, of Course I Can! Deeper Fine-Tuning Attacks That Bypass Token-Level Safety Mechanisms","created":"2025-02-26","cited_by_count":6,"area":"safety"},{"id":"2509.03787","title":"Evaluating the Robustness of Retrieval-Augmented Generation to Adversarial Evidence in the Health Domain","created":"2025-09-04","cited_by_count":6,"area":"rag"},{"id":"2510.10265","title":"Backdoor Collapse: Eliminating Unknown Threats via Known Backdoor Aggregation in Language Models","created":"2025-10-11","cited_by_count":6,"area":"safety"},{"id":"2602.12418","title":"Sparse Autoencoders are Capable LLM Jailbreak Mitigators","created":"2026-02-12","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":"2503.01345","title":"Same Question, Different Words: A Latent Adversarial Framework for Prompt Robustness","created":"2025-03-03","cited_by_count":5,"area":null},{"id":"2406.02619","title":"Unelicitable Backdoors in Language Models via Cryptographic Transformer Circuits","created":"2024-06-03","cited_by_count":4,"area":"safety"},{"id":"2506.16078","title":"Probing the Robustness of Large Language Models Safety to Latent Perturbations","created":"2025-06-19","cited_by_count":4,"area":"alignment"},{"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":"2602.08813","title":"Robust Policy Optimization to Prevent Catastrophic Forgetting","created":"2026-02-09","cited_by_count":3,"area":"alignment"},{"id":"2405.17130","title":"Explaining the role of Intrinsic Dimensionality in Adversarial Training","created":"2024-05-27","cited_by_count":2,"area":null},{"id":"2502.11853","title":"StructTransform: A Scalable Attack Surface for Safety-Aligned Large Language Models","created":"2025-02-17","cited_by_count":2,"area":"alignment"},{"id":"2505.19397","title":"Are Time-Series Foundation Models Deployment-Ready? A Systematic Study of Adversarial Robustness Across Domains","created":"2025-05-26","cited_by_count":2,"area":null},{"id":"2508.09190","title":"Multi-Level Safety Continual Projection for Fine-Tuned Large Language Models without Retraining","created":"2025-08-08","cited_by_count":2,"area":"safety"},{"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":"2602.06911","title":"TamperBench: Systematically Stress-Testing LLM Safety Under Fine-Tuning and Tampering","created":"2026-02-06","cited_by_count":2,"area":"safety"},{"id":"2602.16977","title":"Fail-Closed Alignment for Large Language Models","created":"2026-02-19","cited_by_count":2,"area":"alignment"},{"id":"2502.16366","title":"A Generative Approach to LLM Harmfulness Mitigation with Red Flag Tokens","created":"2025-02-22","cited_by_count":1,"area":null},{"id":"2506.00676","title":"SafeTuneBed: A Toolkit for Benchmarking LLM Safety Alignment in Fine-Tuning","created":"2025-05-31","cited_by_count":1,"area":"alignment"},{"id":"2512.04144","title":"RippleBench: Capturing Ripple Effects Using Existing Knowledge Repositories","created":"2025-12-03","cited_by_count":1,"area":"safety"},{"id":"2601.17037","title":"AMVICC: A Novel Benchmark for Cross-Modal Failure Mode Profiling for VLMs and IGMs","created":"2026-01-20","cited_by_count":1,"area":"lr"},{"id":"2604.12817","title":"Understanding and Improving Continuous Adversarial Training for LLMs via In-context Learning Theory","created":"2026-04-14","cited_by_count":1,"area":"safety"},{"id":"2604.24332","title":"Mitigating Error Amplification in Fast Adversarial Training","created":"2026-04-27","cited_by_count":1,"area":null},{"id":"2605.08496","title":"Latent Personality Alignment: Improving Harmlessness Without Mentioning Harms","created":"2026-05-08","cited_by_count":1,"area":"safety"},{"id":"2605.12813","title":"REALISTA: Realistic Latent Adversarial Attacks that Elicit LLM Hallucinations","created":"2026-05-12","cited_by_count":1,"area":null},{"id":"2607.07918","title":"Efficient Safety Alignment of Language Models via Latent Personality Traits","created":"2026-07-08","cited_by_count":1,"area":"alignment"},{"id":"2607.18454","title":"Estimating Rare Events in Language Models with Proper Evaluation","created":"2026-07-20","cited_by_count":1,"area":null},{"id":"2609.16204","title":"Decoy Direction Optimization: A Post-Hoc Defense Against LLM Abliteration","created":"2026-09-14","cited_by_count":1,"area":"safety"},{"id":"2601.16200","title":"Feature-Space Smoothing: Certified Robustness of Deep Representations","created":"2026-01-22","cited_by_count":0,"area":null},{"id":"2602.15238","title":"Closing the Distribution Gap in Adversarial Training for LLMs","created":"2026-02-16","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":"2604.10403","title":"Latent Instruction Representation Alignment: defending against jailbreaks, backdoors and undesired knowledge in LLMs","created":"2026-04-12","cited_by_count":0,"area":"safety"},{"id":"2604.24350","title":"Unveiling the Backdoor Mechanism Hidden Behind Catastrophic Overfitting in Fast Adversarial Training","created":"2026-04-27","cited_by_count":0,"area":null},{"id":"2606.04168","title":"When Autoregressive Consistency Hurts Safety Alignment","created":"2026-06-02","cited_by_count":0,"area":"alignment"},{"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":"2606.16349","title":"From Refusal Geometry to Safety Geometry: Harmfulness--Refusal Coupling under Dynamic Adversarial Fine-Tuning","created":"2026-06-15","cited_by_count":0,"area":"alignment"},{"id":"2606.26936","title":"Jailbreaking for the Average Jane: Choosing Optimal Jailbreaks via Bandit Algorithms for Automatically Enhanced Queries","created":"2026-06-25","cited_by_count":0,"area":"safety"},{"id":"2607.01239","title":"Breaking Safety at the Token Boundary: How BPE Tokenization Creates Exploitable Gaps in LLM Alignment","created":"2026-05-01","cited_by_count":0,"area":"alignment"},{"id":"2607.28959","title":"Efficient LLM Adversarial Training via Low-Rank Defense and Circuit-Guided Surrogates","created":"2026-07-31","cited_by_count":0,"area":null},{"id":"2608.05045","title":"Gradient Immunity: Null-Space Resistance to Malicious Fine-Tuning","created":"2026-08-05","cited_by_count":0,"area":"alignment"},{"id":"2608.14797","title":"Beyond Tokens: A Survey on Decoding Methods for Large Language and Vision-Language Models","created":"2026-08-14","cited_by_count":0,"area":"mllm"},{"id":"2609.24969","title":"Rare Event Estimation via Iterative Unalignment","created":"2026-09-21","cited_by_count":0,"area":null},{"id":"2609.36862","title":"Safer Content or Firmer Refusals? A Hybrid Perturbation Defense for Alignment under Harmful Fine-tuning","created":"2026-09-29","cited_by_count":0,"area":"alignment"},{"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"},{"id":"2610.13928","title":"AI Models Can Provably Hide Arbitrary Capabilities","created":"2026-10-09","cited_by_count":0,"area":null}],"next_cursor":null}