Skip to content
Home

Current Research · Self-Verifying AI Systems

Self-Verifying AI Systems

When specifications are incomplete, AI systems may make incorrect assumptions and leave users to discover the failures after the fact.

Research question

How can AI systems verify and improve their own work?

Problem

Users can know what they want without being able to specify every requirement in advance. Turning a high-level request into action requires assumptions about unstated constraints, exceptions, and what counts as success. An AI system can produce a plausible result while getting those assumptions wrong.

Publications