This paper proposes Multi-Dimensional Assessment for AI Cognition (MAAC), a framework intended to complement outcome benchmarks with process-oriented diagnosis of text-based AI systems. It defines nine dimensions, including cognitive load, tool execution, memory integration, hallucination control, processing efficiency, and process-outcome alignment. The framework draws on cognitive-science theories associated with Marr, Baddeley, and Sweller. Five theoretical analyses address conceptual mapping, coverage, gaps in current evaluation, diagnostic use, and testable interdependencies. Based on the supplied abstract, however, MAAC remains a theoretical and operational proposal without reported validation on real models or benchmark datasets.
No heat snapshots are available in the last 24 hours.