Community Input

Submit a Research Question

TAR is a general autonomous researcher. Questions across machine learning, AI, and adjacent scientific fields are welcome. Promising submissions are reviewed and queued for investigation. All submissions are read.

Your question has been submitted. Thank you — we’ll review it shortly.

Be as specific as you can. Good questions name a hypothesis, a method, or a gap in current evidence.

Your submission is sent directly to the TAR Research team. Email addresses are used only to reply to your question and are never shared or added to marketing lists.

What makes a good question?

  • Names a specific method, metric, or phenomenon
  • Identifies a gap or contradiction in published results
  • Is testable with a controlled experiment
  • Relates to continual learning or evaluation methodology
  • Is scoped — not “solve continual learning”

TAR is a general researcher

TAR is not limited to a single domain. Current active research includes machine learning and AI, but the system can take up questions in adjacent scientific fields.

  • Machine learning & AI
  • Evaluation methodology
  • Neural network theory
  • Benchmarking & reproducibility
  • Other areas considered

What happens next?

All submissions are reviewed manually. Questions that fall within the current research scope and are testable may be queued for autonomous investigation by the TAR system. We cannot commit to a timeline, but we read everything.

Common Questions

Frequently asked questions

What kinds of questions can I submit?

Any testable question in machine learning, AI, or adjacent scientific fields. Questions that name a specific hypothesis, method, or gap in current evidence are most likely to be taken up.

How long does an investigation take?

Weeks to months, depending on experimental complexity and the current queue. There is no guaranteed timeline — all submissions are reviewed manually.

Will I be notified if my question is selected?

Yes, if you provide an email address. Questions that are not selected will not receive an individual response — but all submissions are read.

Is the research reproducible?

Yes. Every experiment is pre-registered, multi-seed, run under locked conditions, with full provenance records. Results cannot be retrofitted after the fact.

Does TAR only research machine learning?

No. TAR is a general autonomous researcher. Active research focuses on continual learning, but questions in adjacent scientific fields are considered on their merits.

How is TAR different from an AI chatbot?

TAR runs controlled experiments and produces written research with pre-registered criteria. It does not answer questions in conversation. Every published result is tied to a specific experiment, dataset, and set of conditions.