Research note or report

Choose a form before writing: a reading note on someone else's paper, a proposal for an unexecuted study, or a report on work actually performed. Keep the status visible beside the title. This template does not establish that a study was conducted or reviewed.

• Title and reader question: Which decision could this work change?
• Form and status: Reading note, proposal, pilot report, executed study, or research paper. State draft/published and peer-review status separately.
• Authors and contributions: Who did the original work, execution, analysis, and writing? Preserve shared credit.
• Primary source or protocol identity: Public paper URL, DOI/version, or exact protocol revision and artifact hashes. Record what was actually inspected.
• Question and expected observation: What would support the explanation, and what would count against it?
• Background and related work: Which primary sources establish the relevant context? Keep reported results separate from our inference.
• Method and conditions: Task, inputs, environment, model/runtime version, exposed settings, source access, and known substitutions.
• Comparison and baseline: Which conditions differ and which stay matched? Explain confounders and why the baseline is useful.
• Scoring and review: Criteria chosen before execution, scorer identity, review procedure, and unresolved disagreements.
• Cohort and failures: Scheduled and completed attempts, errors, retries, exclusions with reasons, and the unit used for analysis.
• Results: Observed counts and artifacts, including negative results. For a proposal, write “Not executed; no results.”
• Interpretation: Which claim follows from which observation? Label inference and distinguish association from causal evidence.
• Limits and transfer: What was not tested? Which claims about other tasks, models, people, or current products remain unsupported?
• Rights and publication boundary: Which source text, figures, data, outputs, and artifacts may be shared? Keep private supporting material outside the repository.
• Reproduction status: Source inspected, author-reported, locally executed, or independently reproduced, with commands and conditions when applicable.
• Next decision: Adopt a practice within scope, investigate a narrower question, revise the explanation, or leave unresolved.

Use the fields relevant to the form; a short reading note does not need invented experimental sections. A completed report needs enough detail and evidence to inspect how its conclusions were obtained. Follow the research guide and the publication boundary.