auto-review-loop-llm

You can automate the ML research review process by running an autonomous loop that coordinates cross-model critiques, implements fixes, and re-evaluates research artifacts until a target score or round limit is reached.

An autonomous review loop for ML research is an automated cycle that identifies weaknesses in research artifacts, proposes and applies fixes, and re-evaluates the results progressively across multiple model providers.

To set up cross-model critique, you need to configure an MCP server for llm-chat alongside an OpenAI-compatible LLM API, enabling external reviewers to iteratively evaluate and fix your research artifacts.

The autonomous loop persists progress by maintaining a cumulative review log, specifically using REVIEW_STATE.json and AUTO_REVIEW.md files to document traceable round-by-round feedback and improvements.

Yes, you can use alternative LLM APIs. The automated research review loop requires an OpenAI-compatible or alternative LLM API to coordinate cross-model critiques and iterative fixes across providers.

You should avoid using an automated ML research review cycle if your project lacks an accessible llm-chat MCP server or if you do not need structured, traceable round-by-round documentation for progressive artifact improvements.