Published on arXiv Does Not Mean Proven: Read Before Sharing Research News

The phrase 'scientists have proven' is strong; but it should not be used without seeing what the linked document actually says. The fact that a research is available online does not indicate that all its results are independently verified. Especially in news reports spread through preprint servers, it is necessary to examine the status of the publication and the content of the research separately.

arXiv moderation is not referee evaluation

ArXiv's own statementIt makes it clear that content moderation is not a referee review. Acceptance to the server does not mean that a particular magazine is going through the referee process. The opposite is true: seeing a text in arXiv does not necessarily indicate that it has not yet passed through the referee; you need to look at the relevant journal and version information.

This distinction is not meant to devalue research. It is meant to accurately tell the reader what stage of information he encounters. Use 'work published in the journal' statements in accordance with the document with 'the result reported in the preprint'.

Find the question and method instead of the title

What did the research measure, what examples did it use, and what comparison did it make? If you can't find answers to these questions, don't just draw a definitive conclusion from the summary. For example, in a purely fictional AI trial, two models were compared on twenty short texts. From here, one model in all sectors does not come to be superior.

This example is not a real study; it shows how coverage should not be enlarged. The tasks, samples and conditions measured by the research should be preserved in the sentence you will use. Sharing the number or percentage alone does not close the lack of method.

Line between the result and the comment

You can divide your own note into three columns: the finding that the authors reported, the open limit of the study, and my interpretation. Do not confuse these fields. To translate the author's prudent phrase 'may be related' into 'cause found' is to change its meaning, not shorten the text.

Check if these are truly accessible if there are additional data, code, or method details. Finding a link does not mean that you run the code and reproduce the result. Do not describe the test as if you did not.

Check the version and set up a narrow sentence

Read the version of the document and subsequent corrections if available. The screenshot circulating on social media may be from an older version. Verify the most current status from the relevant record if another publication has been moved to.

Briefly state what the work in your sharing shows, what it does not show, and what publication stage it is in. Referee evaluation is also not a guarantee that eliminates every error; however, accurate reporting of the broadcast status helps the reader to make an assessment. Good science news is news that does not hide uncertainty, but describes the strength of the finding along with its limits.