What is Autonomous Science Bench?
The Autonomous Science Bench evaluates AI agents for autonomous scientific discovery. It comprises six design-build-test-learn campaigns.

Illustration: generated for Physis Systems
Short answer: The Autonomous Science Bench is a preprint that has not been peer reviewed, evaluating AI agents for autonomous scientific discovery. It comprises six budgeted design-build-test-learn campaigns. The campaigns are grounded in published experiments across biology, chemistry, and materials science.
The primary source for this article is arXiv cs.AI's Autonomous Science Bench: Evaluating AI Agents for Autonomous Scientific Discovery. It is a preprint, which means it has not yet been peer reviewed.
What is Autonomous Science Bench?
The Autonomous Science Bench evaluates AI agents for autonomous scientific discovery. It comprises six budgeted design-build-test-learn campaigns.
How does it work?
Agents select batches of designs and receive measurements from a lab service. Success requires passing two calibrated gates: a discovery gate and a learning gate.
What are the results?
The best agent-model configuration succeeds in 72% of trials. The hardest campaign is solved in only five of 30 attempts.
What do we know, and what is still open?
What is confirmed: The Autonomous Science Bench evaluates AI agents for autonomous scientific discovery, and the best agent-model configuration succeeds in 72% of trials.
What is still open: The preprint has not been peer reviewed, and the results are based on a limited number of trials.
Why does Physis Systems cover this?
Physis Systems follows research at the frontier, from space to fundamental science. For more stories like this one, visit the Physis Systems blog.
About Physis Systems
Physis Systems (Physis Defense Systems, Inc.) builds Inventi Matrix, an invention engineering platform available in the United States and Canada at inventimatrix.com, and Lumina AI Companion, a personal AI companion you can meet on the Lumina page.
Frequently asked questions
What is the purpose of the Autonomous Science Bench?
The purpose of the Autonomous Science Bench is to evaluate AI agents for autonomous scientific discovery.
How many campaigns are included in the bench?
The bench comprises six budgeted design-build-test-learn campaigns.
What is the success rate of the best agent-model configuration?
The best agent-model configuration succeeds in 72% of trials.


