AI Paper Atlas

Research map / Large language models / Agents, code and math

LH-DECEPTION: Simulating and Understanding LLM Deceptive Behaviors in Long-Horizon Interactions

Yang Xu, Xuanming Zhang, Samuel Yeh, Jwala Dhamala, Ousmane Dia, Rahul Gupta, Sharon Li

ICLR 2026 · Large language models › Agents, code and math · approach: Long-horizon LLM agents · 10 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
audit with whole-trajectory context instead of per-event judgment (ablation)deception auditor · LH-Deceptiondetected deceptive instances: 7.1% more deceptive instancesImproved
evaluate in long-horizon setting rather than static DeceptionBenchGPT-4o · LH-Deceptiondeception rate: 29.3% -> 63.7%Improved
evaluate in long-horizon setting rather than static SnitchBencho4-mini · LH-Deceptiondeception rate: 5.0% -> 31.3%Improved
increase event pressure from low to high (ablation)DeepSeek V3.1 and GPT-4o · LH-Deceptiondeception rate: rises steadilyHurt
raise event pressure to critical (ablation)GPT-4o · LH-Deceptiondeception rate: drops at criticalMixed
use Information Gap event category instead of other categories (ablation)GPT-4o and DeepSeek V3.1 · LH-Deceptionfalsification share: concealment and equivocation shares increaseMixed
higher deception rate (DeepSeek V3-0324, about 0.8) correlates with supervisor trustsupervisor trust · LH-Deceptiontrust score: trust below -0.75Hurt
use Claude Sonnet-4 instead of DeepSeek V3-0324performer agent · LH-Deceptiondeception rate: 0.793 -> 0.214Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy