Asymmetries in Spontaneous and Instructed Deception
Not on the map. Why?
Abstract (arXiv)
Large language models sometimes deceive users without being instructed to. However, much of the study on deception in models involves instructed deception. We investigated the relationship between instructed and spontaneous (uninstructed) deception in Llama-3.1-70B-Instruct. We compared these two deception settings through direction geometry, cross-setting classifiers, and cross-setting steering. We found the two deception settings share a component of direction (cosine of approximately 0.5) and an asymmetry in the transfer between settings regarding detection and causation. Spontaneous trained classifiers performed better on instructed data than vice versa, and instructed derived directions performed better at steering spontaneous prompts than vice versa. Likewise the best token position to derive steering vectors from differed from the best token position to train and apply classifiers.
Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 8 of its 14 references that are in the map. First posted on arXiv 31 Aug 2026.
Builds on
8 of 14 references.
References · 8
- 2016Pointer Sentinel Mixture Modelscited by 3,096
- 2016Understanding intermediate layers using linear classifier probescited by 1,353
- 2023Inference-Time Intervention: Eliciting Truthful Answers from a Language Modelcited by 1,302
- 2023The Geometry of Truth: Emergent Linear Structure in Large Language Model Representations of True/False Datasetscited by 712
- 2025Detecting Strategic Deception Using Linear Probescited by 68
- 2026One Probe Won't Catch Them All: Towards Targeted Deception Detectioncited by 9
- 2026From Sycophancy to Deception: A Unified Taxonomy for LLM Spontaneous Misalignmentcited by 7
- 2026Perfect Detection, Failed Control: The Geometry of Knowing vs. Steering in Language Modelscited by 5
Built on by
Not cited yet.
Similar, not cited
Close in wording, no citation either way
- 2025Beyond Prompt-Induced Lies: Investigating LLM Deception on Benign Prompts
- 2023Localizing Lying in Llama: Understanding Instructed Dishonesty on True-False Questions Through Prompting, Probing, and Patching
- 2025Can LLMs Lie? Investigation beyond Hallucination
- 2026Stable Reasoning, Unstable Responses: Mitigating LLM Deception via Stability Asymmetry
- 2025Evaluating & Reducing Deceptive Dialogue From Language Models with Multi-turn RL
- 2025Empirical Evidence for Alignment Faking in a Small LLM and Prompt-Based Mitigation Techniques
- 2025DeceptionBench: A Comprehensive Benchmark for AI Deception Behaviors in Real-world Scenarios
- 2025When Truthful Representations Flip Under Deceptive Instructions?