Multi-Model Ensemble Design

Design ensemble combining {{model_candidates}} for {{task_type}}.

Step 1: Evaluate individual model strengths on {{benchmark}}
Step 2: Analyze error correlation between models
Step 3: Design routing logic for query distribution
Step 4: Test ensemble configurations (voting, cascading, routing)
Step 5: Optimize for {{optimization_target}} (accuracy, cost, latency)
Step 6: Validate ensemble improvement over best single model

Justify each design decision.