auto-review-loop-llm

Automating iterative paper review uses an LLM reviewer to score and critique research work, applies minimum fixes, and re-reviews until readiness criteria are met. This loop continues up to a configurable maximum number of rounds.

Yes, autonomous research improvement works with any OpenAI-compatible chat API via an llm-chat MCP server. A curl fallback mechanism is also supported to ensure the review loop runs if the primary connection method fails.

State persistence for multi-round LLM critique saves cumulative review logs and compact recovery state after each round. This supports interruption and resumption so you can continue iterative research refinement without losing progress.

Assessing top-venue readiness involves iteratively evaluating claims, methods, results, and weaknesses through automated LLM critiques. The loop stops when the reviewer's score and verdict keywords indicate the work is ready or almost-ready.

You need an llm-chat MCP server to run the auto review loop, though a curl fallback exists if the MCP connection fails. The server bridges the communication between your research draft and the external LLM reviewer.

You should avoid automated research review loops when your work requires subjective human judgment or domain-specific expertise that an LLM cannot accurately evaluate. The tool is designed for structured critique and scoring rather than nuanced qualitative assessment.