mlsanityData to 7 Oct 2026

Trust The Typical

arXiv 2602.04581 ↗First posted 4 Feb 2026cs.CL cs.AI cs.DCcited by 3

On the map: LLM safety and security → Safety alignment, refusal and guardrails

Abstract (arXiv)

Current approaches to LLM safety fundamentally rely on a brittle cat-and-mouse game of identifying and blocking known threats via guardrails. We argue for a fresh approach: robust safety comes not from enumerating what is harmful, but from deeply understanding what is safe. We introduce Trust The Typical (T3), a framework that operationalizes this principle by treating safety as an out-of-distribution (OOD) detection problem. T3 learns the distribution of acceptable prompts in a semantic space and flags any significant deviation as a potential threat. Unlike prior methods, it requires no training on harmful examples, yet achieves state-of-the-art performance across 18 benchmarks spanning toxicity, hate speech, jailbreaking, multilingual harms, and over-refusal, reducing false positive rates by up to 40x relative to specialized safety models. A single model trained only on safe English text transfers effectively to diverse domains and over 14 languages without retraining. Finally, we demonstrate production readiness by integrating a GPU-optimized version into vLLM, enabling continuous guardrailing during token generation with less than 6% overhead even under dense evaluation intervals on large-scale workloads.

Key facts. Cited by 3 papers in the map (data through 7 Oct 2026). Placed in Safety alignment, refusal and guardrails, one of 12 directions in LLM safety and security. Builds on 66 of its 80 references that are in the map. First posted on arXiv 4 Feb 2026.

Builds on

66 of 80 references.

Closest ancestors · 4

Same direction, newest first

  1. 2025Poly-Guard: Massive Multi-Domain Safety Policy-Grounded Guardrail Dataset18 Jun
  2. 2025Enhancing LLM Safety Through a Theoretical Minimax Game Lens7 Feb
  3. 2024WildGuard: Open One-Stop Moderation Tools for Safety Risks, Jailbreaks, and Refusals of LLMs26 Jun
  4. 2023All Languages Matter: On the Multilingual Safety of Large Language Models2 Oct

Other references · 62

  1. 2022Training language models to follow instructions with human feedbackcited by 15,712
  2. 2023Efficient Memory Management for Large Language Model Serving with PagedAttentioncited by 5,767
  3. 2013Intriguing properties of neural networkscited by 5,114
  4. 2017On Calibration of Modern Neural Networkscited by 4,860
  5. 2016Simple and Scalable Predictive Uncertainty Estimation using Deep Ensemblescited by 3,888
  6. 2017Deep reinforcement learning from human preferencescited by 3,846
  7. 2023Qwen Technical Reportcited by 3,273
  8. 2022Constitutional AI: Harmlessness from AI Feedbackcited by 2,950
  9. 2023Universal and Transferable Adversarial Attacks on Aligned Language Modelscited by 2,914
  10. 2020Extracting Training Data from Large Language Modelscited by 2,144
  11. 2022Language Models (Mostly) Know What They Knowcited by 1,783
  12. 2023Jailbroken: How Does LLM Safety Training Fail?cited by 1,565
38 more
  1. 2018A Simple Unified Framework for Detecting Out-of-Distribution Samples and Adversarial Attackscited by 1,472
  2. 2024HarmBench: A Standardized Evaluation Framework for Automated Red Teaming and Robust Refusalcited by 1,330
  3. 2023Jailbreaking Black Box Large Language Models in Twenty Queriescited by 1,326
  4. 2017Enhancing The Reliability of Out-of-distribution Image Detection in Neural Networkscited by 1,325
  5. 2020RealToxicityPrompts: Evaluating Neural Toxic Degeneration in Language Modelscited by 1,235
  6. 2023Llama Guard: LLM-based Input-Output Safeguard for Human-AI Conversationscited by 1,124
  7. 2018Deep Anomaly Detection with Outlier Exposurecited by 939
  8. 2023Semantic Uncertainty: Linguistic Invariances for Uncertainty Estimation in Natural Language Generationcited by 795
  9. 2019How Contextual are Contextualized Word Representations? Comparing the Geometry of BERT, ELMo, and GPT-2 Embeddingscited by 770
  10. 2023Open Problems and Fundamental Limitations of Reinforcement Learning from Human Feedbackcited by 743
  11. 2017Automated Hate Speech Detection and the Problem of Offensive Languagecited by 724
  12. 2023OpenAssistant Conversations -- Democratizing Large Language Model Alignmentcited by 712
  13. 2019Universal Adversarial Triggers for Attacking and Analyzing NLPcited by 709
  14. 2023Baseline Defenses for Adversarial Attacks Against Aligned Language Modelscited by 608
  15. 2023Prompt Injection attack against LLM-integrated Applicationscited by 568
  16. 2022Mind the Gap: Understanding the Modality Gap in Multi-modal Contrastive Representation Learningcited by 553
  17. 2015Towards Open Set Deep Networkscited by 533
  18. 2021ReAct: Out-of-distribution Detection With Rectified Activationscited by 486
  19. 2023Jailbreaking ChatGPT via Prompt Engineering: An Empirical Studycited by 478
  20. 2019Likelihood Ratios for Out-of-Distribution Detectioncited by 451
  21. 2023Catastrophic Jailbreak of Open-source LLMs via Exploiting Generationcited by 439
  22. 2020Generalized ODIN: Detecting Out-of-distribution Image without Learning from Out-of-distribution Datacited by 408
  23. 2022ViM: Out-Of-Distribution with Virtual-logit Matchingcited by 404
  24. 2019Nuanced Metrics for Measuring Unintended Bias with Real Data for Text Classificationcited by 396
  25. 2024LoRA+: Efficient Low Rank Adaptation of Large Modelscited by 365
  26. 2021Exploring the Limits of Out-of-Distribution Detectioncited by 283
  27. 2024SALAD-Bench: A Hierarchical and Comprehensive Safety Benchmark for Large Language Modelscited by 225
  28. 2022Delving into Out-of-Distribution Detection with Vision-Language Representationscited by 215
  29. 2019SemEval-2019 Task 6: Identifying and Categorizing Offensive Language in Social Media (OffensEval)cited by 206
  30. 2022Out-of-Distribution Detection and Selective Generation for Conditional Language Modelscited by 193
  31. 2018Reducing Network Agnostophobiacited by 165
  32. 2020ATOM: Robustifying Out-of-distribution Detection Using Outlier Miningcited by 116
  33. 2020Density of States Estimation for Out-of-Distribution Detectioncited by 74
  34. 2019Towards neural networks that provably know when they don't knowcited by 69
  35. 2021Revisiting Mahalanobis Distance for Transformer-Based Out-of-Domain Detectioncited by 67
  36. 2023Nearest Neighbor Guidance for Out-of-Distribution Detectioncited by 64
  37. 2023Fast Decision Boundary based Out-of-Distribution Detectorcited by 48
  38. 2023Is Fine-tuning Needed? Pre-trained Language Models Are Near Perfect for Out-of-Domain Detectioncited by 29

50 most cited of 62 All 62 (JSON)

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 3.

3 papers.

  1. 2026Path-Lock Expert: Separating Reasoning Mode in Hybrid Thinking via Architecture-Level Separationcited by 3
  2. 2026Privacy Policy Enforcement Guardrails for Data-Sensitive Retrieval-Augmented Generationcited by 0
  3. 2026Send a SCOUT First: Pre-hoc Reasoning for Adaptive Detector Allocation in Prompt-Injection Defensecited by 0