Goal Refinement And Progress Evaluation

via “evaluator-optimizer pattern for iterative output refinement”

Agentic-RAG explores advanced Retrieval-Augmented Generation systems enhanced with AI LLM agents.

Unique: Implements evaluation and optimization as a coupled feedback loop where evaluation results directly drive optimization decisions, rather than treating evaluation as post-hoc validation, enabling continuous quality improvement within the agent execution flow.

vs others: Provides more targeted refinement than simple re-generation by using evaluation feedback to guide optimization, and more efficient than exhaustive search by using LLM reasoning to identify specific improvement opportunities.