Identify the Environment Actually Used
Record the operating system, language or runtime, major libraries, and model versions from the environment that generated the output.
Match Settings and Dependencies
Time references, time zones, coordinates, units, and character encoding affect data input and output. Check dependency versions, configuration files, and run steps together. Randomness or hardware dependence also belongs among the conditions needed for a repeat run.
Check Work Assisted by Generative AI
When reading research that uses generative AI, it matters which tasks the tool assisted and how its output was checked. If generated code is used, a person checks the inputs, units, calculation conditions, and execution results.
See generative AI training and adoption support for help using these tools in research.
Connect Processing to Outputs
Separating acquisition, preprocessing, analysis, and plotting, while linking versions of code, settings, and outputs, makes the effects of changes easier to review. Logs and generation times help trace a figure back to its execution.
Understand What Can Be Shared
Data and software licenses affect what can be shared. Public information may not be sufficient to rerun an analysis under the same conditions.