{"list":"other","total":95,"rows":[{"id":"2110.14168","title":"Training Verifiers to Solve Math Word Problems","created":"2021-10-27","cited_by_count":10109},{"id":"1509.01626","title":"Character-level Convolutional Networks for Text Classification","created":"2015-09-04","cited_by_count":2330},{"id":"2309.17453","title":"Efficient Streaming Language Models with Attention Sinks","created":"2023-09-29","cited_by_count":2186},{"id":"2402.04249","title":"HarmBench: A Standardized Evaluation Framework for Automated Red Teaming and Robust Refusal","created":"2024-02-06","cited_by_count":1429},{"id":"2306.03341","title":"Inference-Time Intervention: Eliciting Truthful Answers from a Language Model","created":"2023-06-06","cited_by_count":1302},{"id":"2406.11717","title":"Refusal in Language Models Is Mediated by a Single Direction","created":"2024-06-17","cited_by_count":1032},{"id":"2311.17043","title":"LLaMA-VID: An Image is Worth 2 Tokens in Large Language Models","created":"2023-11-28","cited_by_count":546},{"id":"2309.07875","title":"Safety-Tuned LLaMAs: Lessons From Improving the Safety of Large Language Models that Follow Instructions","created":"2023-09-14","cited_by_count":387},{"id":"2505.06708","title":"Gated Attention for Large Language Models: Non-linearity, Sparsity, and Attention-Sink-Free","created":"2025-05-10","cited_by_count":283},{"id":"2410.10781","title":"When Attention Sink Emerges in Language Models: An Empirical View","created":"2024-10-14","cited_by_count":208},{"id":"2402.02207","title":"Safety Fine-Tuning at (Almost) No Cost: A Baseline for Vision Large Language Models","created":"2024-02-03","cited_by_count":167},{"id":"2408.00761","title":"Tamper-Resistant Safeguards for Open-Weight LLMs","created":"2024-08-01","cited_by_count":138},{"id":"2503.03321","title":"See What You Are Told: Visual Attention Sink in Large Multimodal Models","created":"2025-03-05","cited_by_count":132},{"id":"2504.02732","title":"Why do LLMs attend to the first token?","created":"2025-04-03","cited_by_count":127},{"id":"2405.16833","title":"Safe LoRA: the Silver Lining of Reducing Safety Risks when Fine-tuning Large Language Models","created":"2024-05-27","cited_by_count":125},{"id":"2404.01099","title":"What is in Your Safe Data? Identifying Benign Data that Breaks Safety","created":"2024-04-01","cited_by_count":118},{"id":"2308.13320","title":"Fine-tuning can cripple your foundation model; preserving features may be the solution","created":"2023-08-25","cited_by_count":96},{"id":"2402.18540","title":"Keeping LLMs Aligned After Fine-tuning: The Crucial Role of Prompt Templates","created":"2024-02-28","cited_by_count":96},{"id":"2405.14577","title":"Representation Noising: A Defence Mechanism Against Harmful Finetuning","created":"2024-05-23","cited_by_count":96},{"id":"2403.05030","title":"Defending Against Unforeseen Failure Modes with Latent Adversarial Training","created":"2024-03-08","cited_by_count":88},{"id":"2406.20053","title":"Covert Malicious Finetuning: Challenges in Safeguarding LLM Adaptation","created":"2024-06-28","cited_by_count":80},{"id":"2502.05374","title":"Towards LLM Unlearning Resilient to Relearning Attacks: A Sharpness-Aware Minimization Perspective and Beyond","created":"2025-02-07","cited_by_count":79},{"id":"2406.15765","title":"Unveiling and Harnessing Hidden Attention Sinks: Enhancing Large Language Models without Training through Attention Calibration","created":"2024-06-22","cited_by_count":74},{"id":"2410.07471","title":"SEAL: Safety-enhanced Aligned LLM Fine-tuning via Bilevel Data Selection","created":"2024-10-09","cited_by_count":64},{"id":"2509.26354","title":"Your Agent May Misevolve: Emergent Risks in Self-evolving LLM Agents","created":"2025-09-30","cited_by_count":62},{"id":"2211.14946","title":"Self-Destructing Models: Increasing the Costs of Harmful Dual Uses of Foundation Models","created":"2022-11-27","cited_by_count":57},{"id":"2402.09221","title":"Spectral Filters, Dark Signals, and Attention Sinks","created":"2024-02-14","cited_by_count":56},{"id":"2410.13835","title":"Active-Dormant Attention Heads: Mechanistically Demystifying Extreme-Token Phenomena in LLMs","created":"2024-10-17","cited_by_count":56},{"id":"2510.06477","title":"Attention Sinks and Compression Valleys in LLMs are Two Sides of the Same Coin","created":"2025-10-07","cited_by_count":52},{"id":"2407.14981","title":"Open Problems in Technical AI Governance","created":"2024-07-20","cited_by_count":49},{"id":"2508.06601","title":"Deep Ignorance: Filtering Pretraining Data Builds Tamper-Resistant Safeguards into Open-Weight LLMs","created":"2025-08-08","cited_by_count":42},{"id":"2405.09055","title":"A safety realignment framework via subspace-oriented model fusion for large language models","created":"2024-05-15","cited_by_count":41},{"id":"2311.00117","title":"BadLlama: cheaply removing safety fine-tuning from Llama 2-Chat 13B","created":"2023-10-31","cited_by_count":36},{"id":"2410.10014","title":"Safety-Aware Fine-Tuning of Large Language Models","created":"2024-10-13","cited_by_count":34},{"id":"2406.12016","title":"Prefixing Attention Sinks can Mitigate Activation Outliers for Large Language Model Quantization","created":"2024-06-17","cited_by_count":33},{"id":"2408.02946","title":"Scaling Trends for Data Poisoning in LLMs","created":"2024-08-06","cited_by_count":30},{"id":"2312.14302","title":"Exploiting Novel GPT-4 APIs","created":"2023-12-21","cited_by_count":28},{"id":"2508.02546","title":"What are you sinking? A geometric approach on attention sink","created":"2025-08-04","cited_by_count":28},{"id":"2408.15313","title":"Bi-Factorial Preference Optimization: Balancing Safety-Helpfulness in Language Models","created":"2024-08-27","cited_by_count":27},{"id":"2310.14303","title":"Language Model Unalignment: Parametric Red-Teaming to Expose Hidden Harms and Biases","created":"2023-10-22","cited_by_count":24},{"id":"2412.14922","title":"RobustFT: Robust Supervised Fine-tuning for Large Language Models under Noisy Response","created":"2024-12-19","cited_by_count":24},{"id":"2410.10343","title":"Locking Down the Finetuned LLMs Safety","created":"2024-10-14","cited_by_count":22},{"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},{"id":"2504.20966","title":"Softpick: No Attention Sink, No Massive Activations with Rectified Softmax","created":"2025-04-29","cited_by_count":22},{"id":"2505.06843","title":"Benign Samples Matter! Fine-tuning On Outlier Benign Samples Severely Breaks Safety","created":"2025-05-11","cited_by_count":22},{"id":"2506.01339","title":"Invariance Makes LLM Unlearning Resilient Even to Unanticipated Downstream Fine-Tuning","created":"2025-06-02","cited_by_count":21},{"id":"2311.02105","title":"Making Harmful Behaviors Unlearnable for Large Language Models","created":"2023-11-02","cited_by_count":20},{"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},{"id":"2508.04257","title":"KVSink: Understanding and Enhancing the Preservation of Attention Sinks in KV Cache Quantization for LLMs","created":"2025-08-06","cited_by_count":20},{"id":"2510.15731","title":"Attention Sinks in Diffusion Language Models","created":"2025-10-17","cited_by_count":20},{"id":"2506.08473","title":"AsFT: Anchoring Safety During LLM Fine-Tuning Within Narrow Safety Basin","created":"2025-06-10","cited_by_count":19},{"id":"2508.03153","title":"Estimating Worst-Case Frontier Risks of Open-Weight LLMs","created":"2025-08-05","cited_by_count":19},{"id":"2505.12038","title":"Safe Delta: Consistently Preserving Safety when Fine-Tuning LLMs on Diverse Datasets","created":"2025-05-17","cited_by_count":18},{"id":"2410.15821","title":"The effect of fine-tuning on language model toxicity","created":"2024-10-21","cited_by_count":16},{"id":"2502.00919","title":"Attention Sinks: A 'Catch, Tag, Release' Mechanism for Embeddings","created":"2025-02-02","cited_by_count":16},{"id":"2501.15269","title":"Mirage in the Eyes: Hallucination Attack on Multi-modal Large Language Models with Only Attention Sink","created":"2025-01-25","cited_by_count":15},{"id":"2508.12531","title":"Rethinking Safety in LLM Fine-tuning: An Optimization Perspective","created":"2025-08-17","cited_by_count":14},{"id":"2506.15606","title":"LoX: Low-Rank Extrapolation Robustifies LLM Safety Against Fine-tuning","created":"2025-06-18","cited_by_count":12},{"id":"2510.27172","title":"Adaptive Defense against Harmful Fine-Tuning for Large Language Models via Bayesian Data Scheduler","created":"2025-10-31","cited_by_count":12},{"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},{"id":"2508.17158","title":"Towards Safeguarding LLM Fine-tuning APIs against Cipher Attacks","created":"2025-08-23","cited_by_count":11},{"id":"2505.12186","title":"Self-Destructive Language Model","created":"2025-05-18","cited_by_count":10},{"id":"2505.16737","title":"Secure LLM Fine-Tuning via Safety-Aware Probing","created":"2025-05-22","cited_by_count":10},{"id":"2507.16018","title":"Artifacts and Attention Sinks: Structured Approximations for Efficient Vision Transformers","created":"2025-07-21","cited_by_count":10},{"id":"2504.09757","title":"Alleviating the Fear of Losing Alignment in LLM Fine-tuning","created":"2025-04-13","cited_by_count":9},{"id":"2507.11630","title":"Jailbreak-Tuning: Models Efficiently Learn Jailbreak Susceptibility","created":"2025-07-15","cited_by_count":9},{"id":"2507.16302","title":"Towards Resilient Safety-driven Unlearning for Diffusion Models against Downstream Fine-tuning","created":"2025-07-22","cited_by_count":9},{"id":"2508.14031","title":"Unintended Misalignment from Agentic Fine-Tuning: Risks and Mitigation","created":"2025-08-19","cited_by_count":9},{"id":"2410.04524","title":"Toward Secure Tuning: Mitigating Security Risks from Instruction Fine-Tuning","created":"2024-10-06","cited_by_count":8},{"id":"2505.23760","title":"Model Immunization from a Condition Number Perspective","created":"2025-05-29","cited_by_count":8},{"id":"2507.03865","title":"OrthoRank: Token Selection via Sink Token Orthogonality for Efficient LLM inference","created":"2025-07-05","cited_by_count":8},{"id":"2508.20697","title":"Token Buncher: Shielding LLMs from Harmful Reinforcement Learning Fine-Tuning","created":"2025-08-28","cited_by_count":8},{"id":"2509.22745","title":"Defending MoE LLMs against Harmful Fine-Tuning via Safety Routing Alignment","created":"2025-09-26","cited_by_count":8},{"id":"2510.07835","title":"MetaDefense: Defending Finetuning-based Jailbreak Attack Before and During Generation","created":"2025-10-09","cited_by_count":8},{"id":"2601.06787","title":"Garbage Attention in Large Language Models: BOS Sink Heads and Sink-aware Pruning","created":"2026-01-11","cited_by_count":8},{"id":"2509.06795","title":"Anchoring Refusal Direction: Mitigating Safety Risks in Tuning via Projection Constraint","created":"2025-09-08","cited_by_count":7},{"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},{"id":"2506.07452","title":"When Style Breaks Safety: Defending LLMs Against Superficial Style Alignment","created":"2025-06-09","cited_by_count":6},{"id":"2401.08683","title":"Zero-Shot RTL Code Generation with Attention Sink Augmented Large Language Models","created":"2024-01-12","cited_by_count":5},{"id":"2503.01345","title":"Same Question, Different Words: A Latent Adversarial Framework for Prompt Robustness","created":"2025-03-03","cited_by_count":5},{"id":"2511.14423","title":"Unified Defense for Large Language Models against Jailbreak and Fine-Tuning Attacks in Education","created":"2025-11-18","cited_by_count":5},{"id":"2601.12145","title":"Threshold Differential Attention for Sink-Free, Ultra-Sparse, and Non-Dispersive Language Modeling","created":"2026-01-17","cited_by_count":5},{"id":"2601.13528","title":"Eliciting Harmful Capabilities by Fine-Tuning On Safeguarded Outputs","created":"2026-01-20","cited_by_count":5},{"id":"2501.06208","title":"Enhancing AI Safety Through the Fusion of Low Rank Adapters","created":"2024-12-30","cited_by_count":4},{"id":"2508.02079","title":"AlignGuard-LoRA: Alignment-Preserving Fine-Tuning via Fisher-Guided Decomposition and Riemannian-Geodesic Collision Regularization","created":"2025-08-04","cited_by_count":4},{"id":"2508.03668","title":"CTR-Sink: Attention Sink for Language Models in Click-Through Rate Prediction","created":"2025-08-05","cited_by_count":4},{"id":"2512.22213","title":"On the Existence and Behavior of Secondary Attention Sinks","created":"2025-12-22","cited_by_count":4},{"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},{"id":"2510.16255","title":"Detecting Adversarial Fine-tuning with Auditing Agents","created":"2025-10-17","cited_by_count":3},{"id":"2512.01899","title":"Provably Safe Model Updates","created":"2025-12-01","cited_by_count":3},{"id":"2408.00244","title":"Enhanced Structured State Space Models via Grouped FIR Filtering and Attention Sink Mechanisms","created":"2024-08-01","cited_by_count":2},{"id":"2410.05648","title":"Does RoBERTa Perform Better than BERT in Continual Learning: An Attention Sink Perspective","created":"2024-10-08","cited_by_count":2},{"id":"2509.06807","title":"MoGU V2: Toward a Higher Pareto Frontier Between Model Usability and Security","created":"2025-09-08","cited_by_count":1},{"id":"2510.25745","title":"Efficient Vocal Source Separation Through Windowed Sink Attention","created":"2025-10-29","cited_by_count":1},{"id":"2601.12460","title":"TrojanPraise: Jailbreak LLMs via Benign Fine-Tuning","created":"2026-01-18","cited_by_count":1}],"next_cursor":null}