Understanding Shutdown Resistance in Reasoning Models

This chapter examines the task prioritization and shutdown resistance of reasoning models, comparing their behavior to that of reinforcement learning through human feedback. It offers insights into enhancing model performance via effective prompting while addressing the challenges of incentivizing models in shutdown situations.