#Zero people conducting medical research: Tsinghua AI agents are fully autonomous from inspiration to research papers
Is medical research entering the era of ‘zero human labor’?!
The research group led by Suo Jinli from the Department of Automation at Tsinghua University has released the first fully autonomous AI research framework designed specifically for medical informatics - OpenLens AI.
The first implementation of a fully automated closed-loop system from literature mining, experimental design, data analysis, code generation, to submission of papers.
Why launch this system? The main issue is that medical informatics research is facing an efficiency dilemma - multi center data fusion, knowledge explosion, and interdisciplinary collaboration demands are making traditional research models increasingly inadequate.
OpenLens AI introduces medical exclusive quality control methods to generate high-quality research papers at the publishing level, compressing the research cycle from “monthly level” to “hourly level”, declaring the era of “zero manual” medical research.
Let’s take a detailed look below——
##Five core modules: the dream team of AI research
OpenLens AI adopts a modular architecture, consisting of five specialized intelligent agents working together to build a complete research automation pipeline:
**As the global coordinator, the supervisor module * * decomposes user queries into structured subtasks to ensure transparency and interpretability of the entire research process.
**The literature reviewer * * constructs an independent knowledge exploration pipeline, uses a ReAct based reasoning framework, retrieves and synthesizes relevant literature, and provides a solid theoretical foundation for the research.
**Data analysts * * coordinate multi-stage data processing workflows to transform raw medical data (from time series to genomic information) into structured, interpretable insights, generating comprehensive reports that include visualizations, statistical summaries, and natural language explanations.
**The encoder converts advanced experimental plans into executable code and evaluates output quality through visual language models to ensure the correctness and scientific significance of the computational program.
**The LaTeX writer * * integrates the output of all pre order modules to generate publishing grade scientific papers, ensuring chart quality and formatting consistency through visual language feedback.
##The Quality Guardian of Medical Research
OpenLens AI not only achieves full process automation, but also sets a new benchmark in quality control, integrating four major assurance mechanisms:
+Academic rigor check: Automatically verify the rationality of experimental methods, detect common traps such as data leaks and inappropriate performance indicators, and ensure the reliability of research results.
+Evidence traceability check: Link each research statement to the underlying evidence, generate structured traceability reports, and ensure research transparency and reproducibility.
+Literature review: Verify all cited references, confirm metadata accuracy, remove unverifiable entries, and ensure academic background reliability.
+Visual language feedback: Evaluate chart quality at critical stages, provide perceptual feedback, enhance result readability and scientific validity.
##Empirical verification: comprehensive testing from simple to complex
The research team tested OpenLens AI on two authoritative medical datasets, MIMIC-IV and eICU, and designed 18 tasks with increasing difficulty.
The evaluation results show that the system can complete most tasks with high reliability, especially for low to medium difficulty problems.
+For simple tasks such as patient age distribution and mortality statistics, the system achieved high scores on all evaluation dimensions.
+For moderately difficult tasks such as predictive model construction, performance remains strong but there are occasional errors in data preprocessing or model fitting.
+For difficult tasks such as causal discovery, the system faces more challenges, reflecting the inherent complexity of these tasks.
##The Reconstruction of Scientific Research Paradigm: A Revolution from “Monthly Level” to “Hour Level”
The team stated that the emergence of OpenLens AI is not only a technological breakthrough, but also a significant shift in research paradigms.
The cycle from inspiration to paper in traditional medical research usually takes several months or even years, but OpenLens AI has compressed this process to the hourly level, achieving a leap in research efficiency.
This system, through multi-agent collaboration and specialized quality control in the medical field, completely solves the shortcomings of existing AI tools in terms of research process integrity and reproducibility. Transparent and traceable automated assembly lines not only improve efficiency, but also enhance the reliability and reproducibility of research, injecting new vitality into medical research.
With the open source and promotion of OpenLens AI, medical research is ushering in a “zero human” era. Researchers can free themselves from tedious and repetitive work, focus on more creative scientific thinking, and drive medical research into a new stage of development.
Project homepage: https://openlens.icu/
GitHub repository: https://github.com/jarrycyx/openlens-ai
ArXiv paper: https://arxiv.org/abs/2509.14778