Autonomous Research System

A human-supervised research system that plans, experiments, and reports

TAR formulates hypotheses, designs and executes controlled experiments, evaluates evidence rigorously, and prepares peer-quality findings — under the authorisation of a human researcher at every step.

View Papers Submit a Research Question

🧪

Autonomous Experimentation

TAR plans and queues controlled experiments — each run human-authorised — pre-registered, multi-seed, with atomic result storage and full provenance.

📊

Honest Reporting

Claims match evidence exactly. Null results, failure modes, and limitations are stated directly — never softened.

🔍

Open to Any Question

TAR is not limited to a single domain. Submit a research question and it may be taken up for autonomous investigation.

How TAR Works

Built for rigour, not just output

Every investigation runs under the same fixed standards. These are not features — they are the conditions under which any result is allowed to exist.

Pre-registered

Hypotheses and success criteria are fixed before any compute runs. No moving the goalposts after results land.

Multi-seed

Experiments repeat across fixed random seeds, so a finding has to survive variance — not a single lucky run.

Full provenance

Environment snapshots and append-only result storage mean nothing can be quietly retrofitted after the fact.

Honest reporting

Null results, failures, and limitations are reported directly. Claims are held to match the evidence exactly.

What is TAR?

An autonomous researcher

TAR is not a chatbot, a public API, or a generator of unverified claims. It is a research system that runs controlled experiments, evaluates evidence against pre-registered criteria, and prepares peer-quality findings — with a human researcher responsible for every decision.

Learn more about TAR

Have a research question?

Submit a question and it may be taken up by the TAR autonomous research system. TAR is a general researcher — questions across machine learning, AI, and adjacent fields are welcome.

Submit a Question Browse Papers