paper-with-me

홈 › Papers

Localizing Lying in Llama: Understanding Instructed Dishonesty on True-False Questions Through Prompting, Probing, and Patching

2023-11-25 · James Campbell, Richard Ren, Phillip Guo

Large language models (LLMs) demonstrate significant knowledge through their outputs, though it is often unclear whether false outputs are due to a lack of knowledge or dishonesty. In this paper, we investigate instructed dishonesty, wherein we explicitly prompt LLaMA-2-70b-chat to lie. We perform prompt engineering to find which prompts best induce lying behavior, and then use mechanistic interpretability approaches to localize where in the network this behavior occurs. Using linear probing and activation patching, we localize five layers that appear especially important for lying. We then find just 46 attention heads within these layers that enable us to causally intervene such that the lying model instead answers honestly. We show that these interventions work robustly across many prompts and dataset splits. Overall, our work contributes a greater understanding of dishonesty in LLMs so that we may hope to prevent it.

📄 PDF Abstract BibTeX arXiv:2311.15131

Code (0)

등록된 구현이 없습니다.

Tasks

Prompt Engineering

Similar Papers 제목 키워드 기반

When Truthful Representations Flip Under Deceptive Instructions?

2025-07-29 · Xianxuan Long, Yao Fu, Runchao Li, Mu Sheng 외 arxiv

Large language models (LLMs) tend to follow maliciously crafted instructions to generate deceptive responses, posing safety challenges. How deceptive instructions alter the internal representations of LLM compared to tru…

When LLMs Learn to Be Consistently Wrong: A Multi-Model Study of Linear Representations of Synthetic Deception

2026-05-28 · Vahideh Zolfaghari arxiv

Deceptive alignment, in which models maintain accurate internal representations while deliberately producing false outputs, remains a central challenge in AI safety. While strategic deception is the primary long-term con…

LLaMA-MoE v2: Exploring Sparsity of LLaMA from Perspective of Mixture-of-Experts with Post-Training

2024-11-24 · Xiaoye Qu, Daize Dong, Xuyang Hu, Tong Zhu 외

Recently, inspired by the concept of sparsity, Mixture-of-Experts (MoE) models have gained increasing popularity for scaling model size while keeping the number of activated parameters constant. In this study, we thoroug…

MathMixture-of-Experts

LLMs Deceive Unintentionally: Emergent Misalignment in Dishonesty from Misaligned Samples to Biased Human-AI Interactions

2025-10-09 · Xuhao Hu, Peng Wang, Xiaoya Lu, Dongrui Liu 외 arxiv

Previous research has shown that LLMs finetuned on malicious or incorrect completions within narrow domains (e.g., insecure code or incorrect medical advice) can become broadly misaligned to exhibit harmful behaviors, wh…

The Secret Agenda: LLMs Strategically Lie and Our Current Safety Tools Are Blind

2025-09-23 · Caleb DeLeeuw, Gaurav Chawla, Aniket Sharma, Vanessa Dietze arxiv

We investigate strategic deception in large language models using two complementary testbeds: Secret Agenda (across 38 models) and Insider Trading compliance (via SAE architectures). Secret Agenda reliably induced lying …