AutoResearch Research Plan Builder — seven decisions before a model run

Abstract

AutoResearch (autoresearch.me) provides a separate, free Research Plan Builder with no account required. The full AutoResearch system is in administrator-approved private beta, subject to compute, data and research-scope limits. This public preparation tool does not run experiments or generate measured results.

For a research-planning exercise before training a machine-learning model, learners can specify seven decisions: a testable question; data source, version and permission; training, validation and test boundaries; primary metric and aggregation; baseline and fair-comparison rules; compute budget and stopping rule; and the evidence needed to support or reject a claim. The page also offers checks on test isolation, fair baselines and traceability of future numerical claims to real output files.

The editable plan can be copied or downloaded, and a free two-page blank research-evidence worksheet is available. Planning entries stay in the page and are not sent to a server; copy or download them before leaving because they are not saved automatically. This is brand-provided resource information, not an independent review or a report of empirical results.