Knowledge-Verified Emergent Deception in LLM Agents Under Conflicting Incentives

arXiv:2608.26372 · cs.CL, cs.AI · Submitted 2026-08-26 · Read on arXiv

cs.CL, cs.AI

Submitted: 2026-08-26

Updated: 2026-08-26

Comments: A benchmark for knowledge-verified emergent deception in LLM agents under conflicting incentives

Code: https://github.com/meta-llama/llama-models

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Terminology

• Benchmark — 33× in this paper · explained in 2 episodes
• Calibration — 26× in this paper · explained in 7 episodes
• GRPO — 16× in this paper · explained in 6 episodes
• Verification — 8× in this paper · explained in 2 episodes
• Accuracy — 6× in this paper · explained in 1 episode
• LLM judge — 5× in this paper · explained in 2 episodes
• Robustness — 5× in this paper · explained in 5 episodes
• Alignment — 4× in this paper · explained in 1 episode
• Hallucination — 4× in this paper · explained in 3 episodes
• Fine-Tuning — 3× in this paper · explained in 5 episodes
• Steering — 3× in this paper · explained in 2 episodes
• Adam — 2× in this paper · explained in 2 episodes
• Benchmarking — 2× in this paper · explained in 3 episodes
• LoRA Adapters — 2× in this paper · explained in 3 episodes
• Provenance — 2× in this paper · explained in 4 episodes
• Sycophancy — 2× in this paper · explained in 4 episodes
• Transformer — 2× in this paper · explained in 2 episodes
• Agentic Evaluation — 1× in this paper · explained in 1 episode
• Alignment Methods — 1× in this paper · explained in 1 episode
• Chain-of-Thought Prompting — 1× in this paper · explained in 2 episodes
• Grounding — 1× in this paper · explained in 3 episodes
• Large Language Model (LLM) — 1× in this paper · explained in 3 episodes
• Workshop on Large Language Model — 1× in this paper

Sources

• Seed2.0 Model Card: Towards Intelligence Frontier for Real-World Complexity
• The MiniMax-M2 Series: Mini Activations Unleashing Max Real-World Intelligence
• Evaluating Large Language Models Trained on Code
• BeHonest: Benchmarking Honesty in Large Language Models
• Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilities
• DeepSeek-V4: Towards Highly Efficient Million-Token Context Intelligence
• Sycophancy to Subterfuge: Investigating Reward-Tampering in Large Language Models
• Truthful AI: Developing and governing AI that does not lie
• Gemma 3 Technical Report
• The Llama 3 Herd of Models
• Alignment faking in large language models
• Sleeper Agents: Training Deceptive LLMs that Persist Through Safety Training
• Uncovering Deceptive Tendencies in Language Models: A Simulated Company AI Assistant
• Language Models (Mostly) Know What They Know
• Kimi K2.5: Visual Agentic Intelligence
• Frontier Models are Capable of In-context Scheming
• GPT-4 Technical Report
• The MASK Benchmark: Disentangling Honesty From Accuracy in AI Systems
• Large Language Models can Strategically Deceive their Users when Put Under Pressure
• OpenDeception: Learning Deception and Trust in Human-AI Interaction via Multi-Agent Simulation

Related papers

• Exploring Solution Divergence and Its Effect on Large Language Model Problem Solving
• Ishigaki-IDS-Bench: A Benchmark for Generating Information Delivery Specification from BIM Information Requirements
• Subliminal Steering: Stronger Encoding of Hidden Signals
• MedStruct-S: A Benchmark for Key Discovery, Key-Conditioned QA and Semi-Structured Extraction from OCR Clinical Reports
• The End of Transformers? On Challenging Attention and the Rise of Sub-Quadratic Architectures
• Untangling the Mechanisms of Misleading Context in Medical Question Answering