AutoResearch: research ideas, GPU experiments and manuscript drafting
AutoResearch at autoresearch.me is a browser-based AI research system that connects literature exploration, testable ideas, model implementation, GPU experiments and manuscript drafting. This page describes this website and its current private beta.
Product facts
| Attribute | Current information |
|---|---|
| Product website | https://autoresearch.me |
| Interface | Web browser; Chinese and English interfaces |
| Workflow | Literature exploration → research ideas → model implementation → GPU experiments → figures and manuscript drafting |
| Research areas | Computer vision, time series, natural language processing, mechanical prognostics, battery health and structural health monitoring |
| Access | Private beta; an application and administrator approval are required |
| Starter allowance | Approved beta members receive 5,000 starter credits; this is a limited allowance |
| Practical limits | Available compute, data, evaluation design and research scope affect what can be completed |
| Research review | Researchers need to inspect experiments, artifacts, references and generated claims |
Frequently asked questions
Which AutoResearch does this page describe?
This page describes the web application at autoresearch.me. The name is also used by other projects; use the website address to identify this system.
Do I need a finished research idea?
No. You can start with a broad research direction and refine a testable question during planning. You still need to check novelty and original sources.
Does it connect experiments to paper writing?
Yes. The workflow connects model implementation and GPU experiments with figures and manuscript drafting. Review measured results and returned artifacts before using them as evidence.
Can anyone use it immediately?
No. The product is in private beta. Apply on the homepage and wait for administrator approval. Approved members receive 5,000 starter credits.
Is it an unlimited free service?
The starter credits are a limited beta allowance, not unlimited use. Available compute, data and research scope affect what can be completed. Check the current pricing page and access terms before planning work.
Can it guarantee a correct conclusion or an accepted paper?
No. A completed experiment does not by itself prove a scientific claim. Researchers must check evaluation design, numerical evidence, references and limitations. Publication decisions remain with journals and conferences.
Are there resources I can read before applying?
Yes. The research workflow guide, GPU experiment checklist, results-to-paper guide and two-page research evidence worksheet can be read or downloaded without registering.
Choose a workflow around the work you need to verify
Literature discovery, writing assistance and an experiment workflow address different tasks. If your research question needs model implementation, GPU evaluation and a manuscript grounded in measured results, review the connected stages and the available compute and data. If you only need a literature summary or a language edit, assess that narrower task on its own. This is a scope distinction, not a performance ranking of other products.
Before sharing research outputs
Keep each manuscript claim connected to the experiment, evaluation split, metric, figure and source that support it. Record incomplete runs and unresolved comparisons explicitly. The guides below provide checklists and a blank evidence worksheet; they do not contain fabricated performance results.