[Paper Review] Kosmos: An AI Scientist for Autonomous Discovery

Ludovico Mitchener, Angela Yiu+35 more|ArXiv.org|Nov 4, 2025

Biomedical Text Mining and Ontologies3 citations

TL;DR

Kosmos automates data-driven scientific discovery via a structured world model that coordinates data analysis and literature search across up to 12-hour cycles, delivering traceable scientific reports and multiple domain discoveries with measurable expert-time savings.

ABSTRACT

Data-driven scientific discovery requires iterative cycles of literature search, hypothesis generation, and data analysis. Substantial progress has been made towards AI agents that can automate scientific research, but all such agents remain limited in the number of actions they can take before losing coherence, thus limiting the depth of their findings. Here we present Kosmos, an AI scientist that automates data-driven discovery. Given an open-ended objective and a dataset, Kosmos runs for up to 12 hours performing cycles of parallel data analysis, literature search, and hypothesis generation before synthesizing discoveries into scientific reports. Unlike prior systems, Kosmos uses a structured world model to share information between a data analysis agent and a literature search agent. The world model enables Kosmos to coherently pursue the specified objective over 200 agent rollouts, collectively executing an average of 42,000 lines of code and reading 1,500 papers per run. Kosmos cites all statements in its reports with code or primary literature, ensuring its reasoning is traceable. Independent scientists found 79.4% of statements in Kosmos reports to be accurate, and collaborators reported that a single 20-cycle Kosmos run performed the equivalent of 6 months of their own research time on average. Furthermore, collaborators reported that the number of valuable scientific findings generated scales linearly with Kosmos cycles (tested up to 20 cycles). We highlight seven discoveries made by Kosmos that span metabolomics, materials science, neuroscience, and statistical genetics. Three discoveries independently reproduce findings from preprinted or unpublished manuscripts that were not accessed by Kosmos at runtime, while four make novel contributions to the scientific literature.

Motivation & Objective

• Motivate autonomous scientific discovery through iterative cycles of data analysis and literature review.
• Enable coherent, multi-agent exploration of open-ended objectives using a shared world model.
• Produce scientifically reportable findings with traceable citations to data and literature.

Proposed method

• Use two Edison Scientific agents (data analysis and literature search) running in parallel.
• Maintain a structured world model to share and synthesize outputs across agents.
• Execute up to ten tasks per cycle and iteratively update the world model.
• Generate three to four scientific reports with statements tightly linked to data or literature.
• Assess report accuracy via expert evaluation of statements, literature sources, and analyses.
• Scale exploration up to 200 agent rollouts with ~42,000 lines of code written per run.

Figure 1: Kosmos workflow and performance. a) Overall workflow for Kosmos. (left) Kosmos is provided with an initial dataset and broad research objective specified by a scientist. (middle) The Kosmos world model coordinates data analysis and literature search agents to identify key discoveries. (rig

Experimental results

Research questions

• RQ1 Can an AI-driven system autonomously conduct iterative data analysis and literature review to fulfill an open-ended research objective?
• RQ2 Does a structured world model enable coherent cross-agent reasoning and citing of sources for traceable scientific reports?
• RQ3 What is the accuracy and expert-time equivalence of Kosmos-generated discoveries across diverse domains?

Key findings

• Kosmos executes an average of 42,000 lines of code and reads about 1,500 papers per run.
• Independent experts judged 79.4% of Kosmos statements as accurate.
• A 20-cycle Kosmos run was estimated by collaborators to equal roughly six months of human research time.
• Kosmos produced seven discoveries spanning metabolomics, materials science, neuroscience, and statistical genetics.
• Three discoveries reproduced unpublished or post-cutoff findings; four contributed novel insights.
• All statements in Kosmos reports are linked to the data analysis outputs or cited literature for traceability.

Figure 2: Kosmos reproduces unpublished discovery in neuroprotection metabolomics. a) Research objective and dataset description provided to Kosmos and experimental design for untargeted metabolomic profiling of mouse brains. b) Barplot showing up-regulation of purine and pyrimidine salvage products