auto-review-loop-llm

An autonomous review loop with an LLM iteratively evaluates research, identifies weaknesses, applies fixes, and re-reviews across multiple rounds until a positive assessment or maximum round cap is reached.

State persistence for LLM research improvement saves progress across rounds by writing persistent logs to a REVIEW_STATE.json file and an AUTO_REVIEW.md file, ensuring auditable tracking of scores and fixes.

Yes, you can use OpenAI-compatible LLMs from multiple providers for research evaluation by configuring them via MCP-based configuration to support diverse models within the same autonomous review loop.

Scoring thresholds guide research submission readiness by defining adjustable metrics within the review loop, dictating when the LLM evaluation achieves a positive assessment and stops iterating.

When an LLM review loop reaches the maximum rounds without a positive assessment, the iterative review and fix cycle terminates, halting autonomous iteration to prevent infinite loops.