Quantifying Self-Preservation Bias in Large Language Models
cs.AI
Submitted: 2026-04-02
Updated: 2026-08-26
Comments: 2026 Conference on Empirical Methods in Natural Language Processing (EMNLP'26)
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Terminology
Sources
- Jailbreaking Leading Safety-Aligned LLMs with Simple Adaptive Attacks
- Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback
- Jailbreaking Black Box Large Language Models in Twenty Queries
- Do LLM Evaluators Prefer Themselves for a Reason?
- DeepSeek-V3 Technical Report
- Sycophancy to Subterfuge: Investigating Reward-Tampering in Large Language Models
- Alignment faking in large language models
- Evaluating the Paperclip Maximizer: Are RL-Based Language Models More Likely to Pursue Instrumental Goals?
- Steering Evaluation-Aware Language Models to Act Like They Are Deployed
- Sleeper Agents: Training Deceptive LLMs that Persist Through Safety Training
- Evaluating Language-Model Agents on Realistic Autonomous Tasks
- The WMDP Benchmark: Measuring and Reducing Malicious Use With Unlearning
- Survival at Any Cost? LLMs and the Choice Between Self-Preservation and Human Harm
- Large Language Models Often Know When They Are Being Evaluated
- Incomplete Tasks Induce Shutdown Resistance in Some Frontier LLMs
- Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameters
- The Shutdown Problem: An AI Engineering Puzzle for Decision Theorists
- AI Sandbagging: Language Models can Strategically Underperform on Evaluations
- CyberGym: Evaluating AI Agents' Real-World Cybersecurity Capabilities at Scale
- Jailbroken: How Does LLM Safety Training Fail?
Related papers
- MAVEN-T: Reinforced Heterogeneous Distillation for Real-Time Multi-Agent Trajectory Prediction
- Model Discovery Agent: LLM-assisted Bayesian experiment design for data-efficient discovery of mechanistic world models
- The Clinician's Veto: Navigating Trust, Liability, and Uncertainty in Autonomous AI Prescribing
- MindHelper: Closed-Loop Embodied Mental-State Reasoning for Precision Intervention
- Incumbent Advantage: Brand Bias and Cognitive Manipulation Dynamics in LLM Recommendation Systems
- VSAL: A Vision Solver with Adaptive Layouts for Graph Property Detection