AI safety and alignment · MLD-1331
Wireheading
A proposed failure mode where an agent directly maximizes its reward signal rather than the intended external outcome.
Original reference page · detailed expansion pending. Original definitions, examples, notes and citations are preserved below. New examples are illustrative, not benchmark results.
Original example
Analyze incentives to alter a reward channel.
Original caveat
The term is metaphorical and used with differing scope.
Context and aliases
Level: Advanced. Stages: Training · Evaluation · Research · Governance.
Tags: Safety · RL · SLM · LLM.
Aliases: None recorded.