Longterm Wiki

AI Evaluation

Methods and frameworks for evaluating AI system safety, capabilities, and alignment properties before deployment, including dangerous capability detection, robustness testing, and deceptive behavior assessment.

Related

Related Pages

Top Related Pages

Risks

Cyberweapons RiskAI Model SteganographyBioweapons Risk

Analysis

AI Safety Intervention Effectiveness MatrixAI Risk Activation Timeline Model

Approaches

Constitutional AIDangerous Capability EvaluationsAI Alignment

Organizations

UK AI Safety InstituteUS AI Safety Institute

Other

Red TeamingDario AmodeiHolden Karnofsky

Policy

Evals-Based Deployment Gates

Concepts

Situational AwarenessSelf-Improvement and Recursive EnhancementAgi DevelopmentGovernance-Focused Worldview

Key Debates

Open vs Closed Source AIGovernment Regulation vs Industry Self-Governance

Tags

evaluationsafety-testingdeployment-decisionscapability-assessmentgovernance