← All drillsFeedback-Loop Patterns
What you'll be able to do- ✓Build an evaluator-optimizer loop, test the judge against planted good and bad answers, and stop on a threshold or a round cap
- ✓Build a critic-actor loop with per-criterion feedback and check in code that each flagged item was really fixed
- ✓Build a self-healing tool caller that diagnoses a failure, changes strategy, and stops on a budget and a repeat guard
- ✓Build a human-in-the-loop run that pauses, saves state, and resumes correctly from approve, correct, redirect, reject and silence
- ✓Choose between the four loops from your own measurements of cost and correctness
Decide how an agent should check and correct its own work before you build it, and know where each feedback loop breaks - the decision that comes before every agent that retries or asks for review.
⌁ The review and retry loops inside coding agents, support-automation approval queues, and LangGraph, Google ADK and AWS Strands agents.
Start this internshipCreate an account to unlock the 6 sections, the workbench, and askThili.
BeginSections
1. Feedback-Loop Patterns: Evaluator-Optimizer, Critic-Actor, Self-Healing, Human-in-the-Loop
🔒 locked2. Lesson 1 - Evaluator-Optimizer: generate, score, improve
🔒 locked3. Lesson 2 - Critic-Actor: critique item by item, fix item by item
🔒 locked4. Lesson 3 - Self-Healing: diagnose the failure, then change strategy
🔒 locked5. Lesson 4 - Human-in-the-Loop: pause, ask, resume
🔒 locked6. Lesson 5 - Choosing a feedback loop
🔒 lockedDig deeper
📄Self-Refine: Iterative Refinement with Self-Feedback (Madaan et al., 2023, arXiv:2303.17651)
paper📄CRITIC: Large Language Models Can Self-Correct with Tool-Interactive Critiquing (Gou et al., 2023, arXiv:2305.11738)
paper📄Teaching Large Language Models to Self-Debug (Chen et al., 2023, arXiv:2304.05128)
paper📄Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena (Zheng et al., 2023, arXiv:2306.05685)
paper🔗Building Effective Agents - Anthropic Engineering
article🔗Groq - free API keys for open models
docs