{"total":155,"rows":[{"id":"2604.02022","title":"ATBench: A Diverse and Realistic Agent Trajectory Benchmark for Safety Evaluation and Diagnosis","created":"2026-04-02","cited_by_count":36,"area":"agents"},{"id":"2601.01592","title":"OpenRT: An Open-Source Red Teaming Framework for Multimodal LLMs","created":"2026-01-04","cited_by_count":21,"area":"mllm"},{"id":"2601.15588","title":"YuFeng-XGuard: A Reasoning-Centric, Interpretable, and Flexible Guardrail Model for Large Language Models","created":"2026-01-22","cited_by_count":19,"area":"lr"},{"id":"2604.02947","title":"AgentHazard: A Benchmark for Evaluating Harmful Behavior in Computer-Use Agents","created":"2026-04-03","cited_by_count":18,"area":"agents"},{"id":"2601.10955","title":"Beyond Max Tokens: Stealthy Resource Amplification via Tool Calling Chains in LLM Agents","created":"2026-01-16","cited_by_count":14,"area":"agents"},{"id":"2512.21008","title":"GateBreaker: Gate-Guided Attacks on Mixture-of-Expert LLMs","created":"2025-12-24","cited_by_count":11,"area":"safety"},{"id":"2602.20880","title":"When Safety Collides: Resolving Multi-Category Harmful Conflicts in Text-to-Image Diffusion via Adaptive Safety Guidance","created":"2026-02-24","cited_by_count":11,"area":"diffusion"},{"id":"2601.06663","title":"SafePro: Evaluating the Safety of Professional-Level AI Agents","created":"2026-01-10","cited_by_count":10,"area":"agents"},{"id":"2602.03224","title":"TAME: A Trustworthy Test-Time Evolution of Agent Memory with Systematic Benchmarking","created":"2026-02-03","cited_by_count":10,"area":"agents"},{"id":"2602.20708","title":"ICON: Indirect Prompt Injection Defense for Agents based on Inference-Time Correction","created":"2026-02-24","cited_by_count":10,"area":"agents"},{"id":"2603.03205","title":"Learning When to Act or Refuse: Guarding Agentic Reasoning Models for Safe Multi-Step Tool Use","created":"2026-03-03","cited_by_count":8,"area":"lr"},{"id":"2605.01687","title":"MultiBreak: A Scalable and Diverse Multi-turn Jailbreak Benchmark for Evaluating LLM Safety","created":"2026-05-03","cited_by_count":8,"area":"safety"},{"id":"2605.11882","title":"On-Policy Self-Evolution via Failure Trajectories for Agentic Safety Alignment","created":"2026-05-12","cited_by_count":8,"area":"agents"},{"id":"2605.29801","title":"AgentDoG 1.5: A Lightweight and Scalable Alignment Framework for AI Agent Safety and Security","created":"2026-05-28","cited_by_count":8,"area":"agents"},{"id":"2512.01353","title":"The Trojan Knowledge: Bypassing Commercial LLM Guardrails via Harmless Prompt Weaving and Adaptive Tree Search","created":"2025-12-01","cited_by_count":7,"area":"safety"},{"id":"2601.03868","title":"What Matters For Safety Alignment?","created":"2026-01-07","cited_by_count":7,"area":"alignment"},{"id":"2607.01793","title":"Safety Testing LLM Agents at Scale: From Risk Discovery to Evidence-Grounded Verification","created":"2026-07-02","cited_by_count":7,"area":"agents"},{"id":"2602.22554","title":"Multilingual Safety Alignment Via Sparse Weight Editing","created":"2026-02-26","cited_by_count":6,"area":"alignment"},{"id":"2604.14858","title":"Benchmarks for Trajectory Safety Evaluation and Diagnosis in OpenClaw and Codex: ATBench-Claw and ATBench-Codex","created":"2026-04-16","cited_by_count":6,"area":null},{"id":"2605.02946","title":"RouteHijack: Routing-Aware Attack on Mixture-of-Experts LLMs","created":"2026-05-01","cited_by_count":6,"area":"alignment"},{"id":"2606.14502","title":"From Chatbot to Digital Colleague: The Paradigm Shift Toward Persistent Autonomous AI","created":"2026-06-12","cited_by_count":6,"area":null},{"id":"2606.22873","title":"SingGuard: A Policy-Adaptive Multimodal LLM Guardrail with Dynamic Reasoning","created":"2026-06-22","cited_by_count":6,"area":"lr"},{"id":"2607.19829","title":"DARWIN: Evolving Jailbreak Adversary and Guardrail for LLM Safety Evaluation and Protection","created":"2026-07-22","cited_by_count":6,"area":"safety"},{"id":"2608.24777","title":"StepGuard: Learning Step-Level Guardrails with Scalable Supervision and Safety-Utility Balancing","created":"2026-08-25","cited_by_count":6,"area":"agents"},{"id":"2605.17329","title":"LPG: Balancing Efficiency and Policy Reasoning in Latent Policy Guardrails","created":"2026-05-17","cited_by_count":5,"area":"lr"},{"id":"2511.20994","title":"GuardTrace-VL: Detecting Unsafe Multimodel Reasoning via Iterative Safety Supervision","created":"2025-11-26","cited_by_count":4,"area":"lr"},{"id":"2602.01600","title":"Expected Harm: Rethinking Safety Evaluation of (Mis)Aligned LLMs","created":"2026-02-02","cited_by_count":4,"area":"lr"},{"id":"2603.02219","title":"NExT-Guard: Training-Free Streaming Safeguard without Token-Level Labels","created":"2026-02-10","cited_by_count":4,"area":"mechinterp"},{"id":"2605.31140","title":"EvoDefense: Co-Evolving Black-Box Defense with Large Language Models","created":"2026-05-29","cited_by_count":4,"area":null},{"id":"2606.01166","title":"BraveGuard: From Open-World Threats to Safer Computer-Use Agents","created":"2026-05-31","cited_by_count":4,"area":"agents"},{"id":"2510.19169","title":"OpenGuardrails: A Configurable, Unified, and Scalable Guardrails Platform for Large Language Models","created":"2025-10-22","cited_by_count":3,"area":"safety"},{"id":"2510.21190","title":"The Trojan Example: Jailbreaking LLMs through Template Filling and Unsafety Reasoning","created":"2025-10-24","cited_by_count":3,"area":"alignment"},{"id":"2601.05739","title":"PII-VisBench: Evaluating Personally Identifiable Information Safety in Vision Language Models Along a Continuum of Visibility","created":"2026-01-09","cited_by_count":3,"area":"mllm"},{"id":"2601.13655","title":"Why Does the LLM Stop Computing: An Empirical Study of User-Reported Failures in Open-Source LLMs","created":"2026-01-20","cited_by_count":3,"area":null},{"id":"2601.19231","title":"LLMs Can Unlearn Refusal with Only 1,000 Benign Samples","created":"2026-01-27","cited_by_count":3,"area":"alignment"},{"id":"2602.04195","title":"Semantic Consensus Decoding: Backdoor Defense for Verilog Code Generation","created":"2026-02-04","cited_by_count":3,"area":"safety"},{"id":"2602.11096","title":"Safety Recovery in Reasoning Models Is Only a Few Early Steering Steps Away","created":"2026-02-11","cited_by_count":3,"area":"lr"},{"id":"2603.04469","title":"Cross-Layer Semantic Flow Reconstruction for Attack Detection in Agentic Systems","created":"2026-03-04","cited_by_count":3,"area":"agents"},{"id":"2604.08005","title":"Preference Redirection via Attention Concentration: An Attack on Computer Use Agents","created":"2026-04-09","cited_by_count":3,"area":"agents"},{"id":"2605.08778","title":"Not All Turns Matter: Credit Assignment for Multi-Turn Jailbreaking","created":"2026-05-09","cited_by_count":3,"area":"safety"},{"id":"2605.08930","title":"Internalizing Safety Understanding in Large Reasoning Models via Verification","created":"2026-05-09","cited_by_count":3,"area":"lr"},{"id":"2605.11002","title":"MT-JailBench: A Modular Benchmark for Understanding Multi-Turn Jailbreak Attacks","created":"2026-05-10","cited_by_count":3,"area":"safety"},{"id":"2605.16346","title":"PropGuard: Safeguarding LLM-MAS via Propagation-Aware Exploration and Remediation","created":"2026-05-08","cited_by_count":3,"area":"agents"},{"id":"2606.01991","title":"SafeMCP: Proactive Power Regulation for LLM Agent Defense via Environment-Grounded Look-Ahead Reasoning","created":"2026-06-01","cited_by_count":3,"area":"lr"},{"id":"2606.25013","title":"Do Thinking Tokens Help with Safety?","created":"2026-06-23","cited_by_count":3,"area":"lr"},{"id":"2601.01739","title":"K-EXAONE Technical Report","created":"2026-01-05","cited_by_count":2,"area":null},{"id":"2601.22737","title":"Lingua-SafetyBench: A Benchmark for Safety Evaluation of Multilingual Vision-Language Models","created":"2026-01-30","cited_by_count":2,"area":"mllm"},{"id":"2602.07954","title":"Bielik Guard: Efficient Polish Language Safety Classifiers for LLM Content Moderation","created":"2026-02-08","cited_by_count":2,"area":null},{"id":"2602.10179","title":"When the Prompt Becomes Visual: Vision-Centric Jailbreak Attacks for Large Image Editing Models","created":"2026-02-10","cited_by_count":2,"area":"safety"},{"id":"2602.13562","title":"Mitigating the Safety-utility Trade-off in LLM Alignment via Adaptive Safe Context Learning","created":"2026-02-14","cited_by_count":2,"area":"lr"},{"id":"2602.16835","title":"NeST: Neuron Selective Tuning for LLM Safety","created":"2026-02-18","cited_by_count":2,"area":"alignment"},{"id":"2604.12616","title":"Every Picture Tells a Dangerous Story: Memory-Augmented Multi-Agent Jailbreak Attacks on VLMs","created":"2026-04-14","cited_by_count":2,"area":"mllm"},{"id":"2605.22321","title":"ASEval: Automated Trajectory-Level Security Testing for Autonomous Agents","created":"2026-05-21","cited_by_count":2,"area":"agents"},{"id":"2605.27489","title":"HARP: Measuring Harm Amplification in Multi-Agent LLM Systems","created":"2026-05-26","cited_by_count":2,"area":"agents"},{"id":"2607.03069","title":"SafeGuard: A Multi-Agent Perception-Reasoning Framework for Social-Risk AI-Generated Video Detection","created":"2026-07-03","cited_by_count":2,"area":"lr"},{"id":"2607.13081","title":"SingGuard-NSFA: Extensible Guardrails for Agentic AI via Generative Reasoning and Real-Time Classification","created":"2026-07-13","cited_by_count":2,"area":"agents"},{"id":"2511.19009","title":"Understanding and Mitigating Over-refusal for Large Language Models via Representation Intervention","created":"2025-11-24","cited_by_count":1,"area":"safety"},{"id":"2511.19171","title":"Can LLMs Threaten Human Survival? Benchmarking Potential Existential Threats from LLMs via Prefix Completion","created":"2025-11-24","cited_by_count":1,"area":"safety"},{"id":"2602.07107","title":"ShallowJail: Steering Jailbreaks against Large Language Models","created":"2026-02-06","cited_by_count":1,"area":"safety"},{"id":"2603.23171","title":"Adaptively Robust LLM Monitoring via Activation Watermarking","created":"2026-03-24","cited_by_count":1,"area":"safety"},{"id":"2603.25176","title":"Prompt Attack Detection with LLM-as-a-Judge and Mixture-of-Models","created":"2026-03-26","cited_by_count":1,"area":"safety"},{"id":"2603.29038","title":"Trojan-Speak: Bypassing Constitutional Classifiers with No Jailbreak Tax via Adversarial Finetuning","created":"2026-03-30","cited_by_count":1,"area":"safety"},{"id":"2604.16542","title":"TWGuard: A Case Study of LLM Safety Guardrails for Localized Linguistic Contexts","created":"2026-04-17","cited_by_count":1,"area":"safety"},{"id":"2605.00689","title":"BabelSafe: A Policy-Grounded Multilingual Safety Benchmark for LLMs","created":"2026-05-01","cited_by_count":1,"area":"safety"},{"id":"2605.06230","title":"Safactory: A Scalable Agentic Infrastructure for Training Trustworthy Autonomous Intelligence","created":"2026-05-07","cited_by_count":1,"area":"agents"},{"id":"2605.08936","title":"Self-ReSET: Learning to Self-Recover from Unsafe Reasoning Trajectories","created":"2026-05-09","cited_by_count":1,"area":"lr"},{"id":"2605.29068","title":"Robust and Efficient Guardrails with Latent Reasoning","created":"2026-05-27","cited_by_count":1,"area":"lr"},{"id":"2605.30693","title":"Triaging Threats to Specialized Guardrails","created":"2026-05-29","cited_by_count":1,"area":"safety"},{"id":"2606.02640","title":"D-Judge: Disrupting Multi-Turn Jailbreaks using Semantics-Preserving Output Rewriting","created":"2026-05-31","cited_by_count":1,"area":"safety"},{"id":"2606.11409","title":"Risk Under Pressure: Compute-Aware Evaluation of Adversarial Robustness in Language Models","created":"2026-06-09","cited_by_count":1,"area":"safety"},{"id":"2606.19168","title":"Beyond Safe Data: Pretraining-Stage Alignment with Regular Safety Reflection","created":"2026-06-17","cited_by_count":1,"area":"alignment"},{"id":"2606.25442","title":"PolicyAlign: Direct Policy-Based Safety Alignment for Large Language Models","created":"2026-06-24","cited_by_count":1,"area":"alignment"},{"id":"2606.29887","title":"SafePyramid: A Hierarchical Benchmark for In-context Policy Guardrailing","created":"2026-06-29","cited_by_count":1,"area":"safety"},{"id":"2607.18056","title":"An Early Warning of Emerging Biosecurity Risks in Frontier LLMs","created":"2026-07-20","cited_by_count":1,"area":"safety"},{"id":"2607.19913","title":"JANUS: Foreseeing Latent Risk for Long-Horizon Agent Safety","created":"2026-07-22","cited_by_count":1,"area":null},{"id":"2607.21401","title":"When Are Reasoning-Based Guardrails Not Efficient? ResponseGuard: A Fast Vision-Language Guard for Real-Time Moderation","created":"2026-07-23","cited_by_count":1,"area":"lr"},{"id":"2607.25857","title":"Shieldstral","created":"2026-07-28","cited_by_count":1,"area":null},{"id":"2608.07430","title":"Diffusion LLMs as Targets and Adversaries: Mechanistic Safety Exploits","created":"2026-08-07","cited_by_count":1,"area":"safety"},{"id":"2608.23034","title":"ST$^2$U: Stateful Test-Time Unlearning via Restricted Knowledge Boundary Control","created":"2026-08-24","cited_by_count":1,"area":"alignment"},{"id":"2610.00400","title":"Representation Transitions Reveal Emerging Safety Risks in Multi-Turn LLM Agents","created":"2026-09-30","cited_by_count":1,"area":"agents"},{"id":"2512.13703","title":"Safe2Harm: Semantic Isomorphism Attacks for Jailbreaking Large Language Models","created":"2025-12-05","cited_by_count":0,"area":"safety"},{"id":"2601.04199","title":"The Forgotten Shield: Safety Grafting in Parameter-Space for Medical MLLMs","created":"2025-12-05","cited_by_count":0,"area":"mllm"},{"id":"2601.10804","title":"BYOL: Bring Your Own Language Into LLMs","created":"2026-01-15","cited_by_count":0,"area":null},{"id":"2602.06650","title":"Beyond Static Alignment: Hierarchical Policy Control for LLM Safety via Risk-Aware Chain-of-Thought","created":"2026-02-06","cited_by_count":0,"area":"lr"},{"id":"2603.10091","title":"Multi-Stream Perturbation Attack: Breaking Safety Alignment of Thinking LLMs Through Concurrent Task Interference","created":"2026-03-10","cited_by_count":0,"area":"lr"},{"id":"2603.15647","title":"Steering Frozen LLMs: Adaptive Social Alignment via Online Prompt Routing","created":"2026-03-04","cited_by_count":0,"area":"alignment"},{"id":"2603.16192","title":"Structured Semantic Cloaking for Jailbreak Attacks on Large Language Models","created":"2026-03-17","cited_by_count":0,"area":"safety"},{"id":"2603.26511","title":"AMALIA Technical Report: A Fully Open Source Large Language Model for European Portuguese","created":"2026-03-27","cited_by_count":0,"area":null},{"id":"2604.08846","title":"Dictionary-Aligned Concept Control for Safeguarding Multimodal LLMs","created":"2026-04-10","cited_by_count":0,"area":"mllm"},{"id":"2604.08881","title":"Targeted Interpretable Safety Neuron Enhancement for Multilingual Vision-Language Large Models","created":"2026-04-10","cited_by_count":0,"area":"alignment"},{"id":"2604.25716","title":"Cross-Lingual Jailbreak Detection via Semantic Codebooks","created":"2026-04-28","cited_by_count":0,"area":"safety"},{"id":"2605.05630","title":"One Turn Too Late: Learning When to Intervene Against Multi-Turn Malicious Intent","created":"2026-05-07","cited_by_count":0,"area":null},{"id":"2605.08878","title":"Why Do Aligned LLMs Remain Jailbreakable: Refusal-Escape Directions, Operator-Level Sources, and Safety-Utility Trade-off","created":"2026-05-09","cited_by_count":0,"area":"alignment"},{"id":"2605.12265","title":"How Useful Is Cross-Domain Generalization for Training LLM Monitors?","created":"2026-05-12","cited_by_count":0,"area":null},{"id":"2605.13411","title":"Model-Agnostic Lifelong LLM Safety via Externalized Attack-Defense Co-Evolution","created":"2026-05-13","cited_by_count":0,"area":"safety"},{"id":"2605.20654","title":"REFLECTOR: Internalizing Step-wise Reflection against Indirect Jailbreak","created":"2026-05-20","cited_by_count":0,"area":"alignment"},{"id":"2605.22373","title":"Boundary-targeted Membership Inference Attacks on Safety Classifiers","created":"2026-05-21","cited_by_count":0,"area":"safety"},{"id":"2605.23974","title":"AERIC: Anticipatory Hidden-State Monitoring for Implicit Harmful Dialogue","created":"2026-05-13","cited_by_count":0,"area":null},{"id":"2606.09038","title":"Personalization Meets Safety:Mechanisms,Risks,and Mitigations in Personalized LLMs","created":"2026-06-08","cited_by_count":0,"area":"moe"},{"id":"2606.13737","title":"FreoStream:Enhancing Stream Guardrails via Future-Aware Reasoning and Safety-Aligned Optimization","created":"2026-06-11","cited_by_count":0,"area":"safety"}],"next_cursor":"100"}