mlsanityData to 14 Oct 2026

Pragmatic Attack Surface: Vulnerabilities of Implicit Context in Large Language Models

arXiv 2608.09551 ↗First posted 10 Aug 2026cs.CLcited by 0

On the map: LLM alignment / RLHF / preference optimisation → Safety alignment and jailbreak attacks

also in 1 more
Abstract (arXiv)

In the era of large language models (LLMs), attackers often manipulate natural language to elicit unsafe or harmful outputs, creating a new natural language attack surface unique to LLM-based systems, where attacks directly exploit explicit linguistic cues in user prompts to bypass the safety mechanism of LLMs. However, such attacks can often be mitigated by existing safety alignment algorithms. On the other hand, human language is inherently grounded in pragmatics, necessitating typical context to interpret language, e.g., world knowledge, social norms. However, such contexts are often implicit because they are not directly expressed in human language and are not sufficiently leveraged in safety alignment, creating a fundamental mismatch between human language interpretation and safety alignment approaches. In this paper, we demonstrate that this mismatch exposes vulnerabilities in LLMs. We refer to this vulnerability as the pragmatic attack surface, which can be exploited to achieve high attack success rates. The experimental results demonstrate that our proposed approach outperforms baseline attack methods across various open-source and closed-source models by a substantial margin.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Safety alignment and jailbreak attacks, one of 10 directions in LLM alignment / RLHF / preference optimisation. Builds on 33 of its 53 references that are in the map. First posted on arXiv 10 Aug 2026.

Builds on

33 of 53 references.

References · 33

  1. 2022Training language models to follow instructions with human feedbackcited by 15,971
  2. 2023Direct Preference Optimization: Your Language Model is Secretly a Reward Modelcited by 8,065
  3. 2021On the Opportunities and Risks of Foundation Modelscited by 4,033
  4. 2017Deep reinforcement learning from human preferencescited by 3,912
  5. 2025Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilitiescited by 3,796
  6. 2024HarmBench: A Standardized Evaluation Framework for Automated Red Teaming and Robust Refusalcited by 1,429
  7. 2023Jailbreaking Black Box Large Language Models in Twenty Queriescited by 1,420
  8. 2023Llama Guard: LLM-based Input-Output Safeguard for Human-AI Conversationscited by 1,207
  9. 2021BBQ: A Hand-Built Bias Benchmark for Question Answeringcited by 701
  10. 2024Great, Now Write an Article About That: The Crescendo Multi-Turn LLM Jailbreak Attackcited by 390
  11. 2023NeMo Guardrails: A Toolkit for Controllable and Safe LLM Applications with Programmable Railscited by 365
  12. 2022A Holistic Approach to Undesired Content Detection in the Real Worldcited by 356
21 more
  1. 2024Deliberative Alignment: Reasoning Enables Safer Language Modelscited by 308
  2. 2022Neural Theory-of-Mind? On the Limits of Social Intelligence in Large LMscited by 220
  3. 2021Latent Hatred: A Benchmark for Understanding Implicit Hate Speechcited by 217
  4. 2025SafeChain: Safety of Language Models with Long Chain-of-Thought Reasoning Capabilitiescited by 136
  5. 2025X-Teaming: Multi-Turn Jailbreaks and Defenses with Adaptive Multi-Agentscited by 125
  6. 2024LLMs know their vulnerabilities: Uncover Safety Gaps through Natural Distribution Shiftscited by 86
  7. 2020Are Natural Language Inference Models IMPPRESsive? Learning IMPlicature and PRESuppositioncited by 77
  8. 2022CREPE: Open-Domain Question Answering with False Presuppositionscited by 72
  9. 2022A fine-grained comparison of pragmatic language understanding in humans and language modelscited by 70
  10. 2024PUB: A Pragmatics Understanding Benchmark for Assessing LLMs' Pragmatics Capabilitiescited by 59
  11. 2025PolyGuard: A Multilingual Safety Moderation Tool for 17 Languagescited by 59
  12. 2022The Goldilocks of Pragmatic Understanding: Fine-Tuning Strategy Matters for Implicature Resolution by LLMscited by 58
  13. 2021NOPE: A Corpus of Naturally-Occurring Presuppositions in Englishcited by 23
  14. 2025Pragmatic Inference Chain (PIC) Improving LLMs' Reasoning of Authentic Implicit Toxic Languagecited by 8
  15. 2025NEXUS: Network Exploration for eXploiting Unsafe Sequences in Multi-Turn LLM Jailbreakscited by 7
  16. 2025Diagnosing Moral Reasoning Acquisition in Language Models: Pragmatics and Generalizationcited by 6
  17. 2025Learning to Conceal Risk: Controllable Multi-turn Red Teaming for LLMs in the Financial Domaincited by 6
  18. 2026The Echo Chamber Multi-Turn LLM Jailbreakcited by 4
  19. 2025Diagnosing the Performance Trade-off in Moral Alignment: A Case Study on Gender Stereotypescited by 2
  20. 2025From Training to Generalization: Improving Moral Reasoning Through Pragmatic Inferencecited by 2
  21. 2026Unveiling the Limits of Large Language Models in Inferring Pragmatic Meaning from Non-Verbal Responsescited by 1

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2025Beyond Context: Large Language Models' Failure to Grasp Users' Intent
  2. 2024Emulated Disalignment: Safety Alignment for Large Language Models May Backfire!
  3. 2023Survey of Vulnerabilities in Large Language Models Revealed by Adversarial Attacks
  4. 2026IndicSafeEval: Safety Robustness of Large Language Models under Multilingual Persuasive Jailbreak Attacks
  5. 2024Safety Alignment Should Be Made More Than Just a Few Tokens Deep
  6. 2024The Dark Side of Human Feedback: Poisoning Large Language Models via User Inputs
  7. 2025SABER: Uncovering Vulnerabilities in Safety Alignment via Cross-Layer Residual Connection
  8. 2026LASA: Language-Agnostic Semantic Alignment at the Semantic Bottleneck for LLM Safety