Concept-Based Intervenability in AI models leverages intermediate representations aligned with human-understandable concepts as the primary interface for interaction, often utilizing a 'concept bottleneck' layer to channel reasoning through these concepts.

Authors

• Person: Not available Organization: Springer
  A Comprehensive Review of Neuro-symbolic AI for Robustness ...

Sources

• A Comprehensive Review of Neuro-symbolic AI for Robustness ... link.springer.com Springer via serper

Referenced by nodes (1)

• AI models concept