Does a preprint include revisions made in response to peer review?

I am using Elsevier as example since they are often picked on for many of their policies. Elsevier attempts to be clear about their sharing policy and allows both preprints and accepted manuscripts to be shared. In "Elsevier speak" (which may not agree with standard usage of the terms in academia)

Preprint

This is the author's own write-up of research results and analysis that has not been peer reviewed, nor had any other value added to it by a publisher (such as formatting, copy-editing, technical enhancements, and the like).

Accepted manuscript

An accepted manuscript is the manuscript of an article that has been accepted for publication and which typically includes author-incorporated changes suggested during submission, peer review, and editor-author communications. They do not include other publisher value-added contributions such as copy-editing, formatting, technical enhancements and (if relevant) pagination.

Elsevier allows for "accepted manuscripts" to be made available via "preprint" repositories (e.g., ArXiv) which of course speaks to how confusing the terminology is. So at least for articles published by Elsevier the difference between the published version and the version in a preprint repository might be limited to "copy-editing, formatting, technical enhancements".

That said, and one of my pet peeves about preprint repositories, is that the "preprint" may have any number of non-peer-review changes. Maybe during the review process some controversial data or conclusions were removed from the manuscript. There is nothing preventing the authors from re-inserting this into the "preprint".

There is no consistent terminolgoy in this respect. Sherpa/Romeo, a database for publisher’s pre-print policies and probably as close to an authority as you can get on this matter, writes about this:

The terms pre-print and post-print are used to mean different things by different people. This can cause some confusion and ambiguity.

One usage of the term pre-print is to describe the first draft of the article - before peer-review, even before any contact with a publisher. This use is common amongst academics for whom the key modification of an article is the peer-review process.

Another use of the term pre-print is for the finished article, reviewed and amended, ready and accepted for publication - but separate from the version that is type-set or formatted by the publisher. This use is more common amongst publishers, for whom the final and significant stage of modification to an article is the arrangement of the material for putting to print.

[…]

To try to clarify the situation, this listing characterises pre-prints as being the version of the paper before peer review and post-prints as being the version of the paper after peer-review, with revisions having been made.

Publisher’s policies on what authors may post on a preprint server vary a lot and Sherpa/Romeo maintains an overview over these policies. However, authors may not adhere to these policies or make the most of them when publishing their papers. Thus it is usually not possible to tell which version of the paper is published on a preprint repository.

When you want to publish your own papers, I recommend to check both, Sherpa/Romeo and the copyright agreement – most of them are suprisingly human-readable.

The term "pre-print" generally refers to the author's final draft of the manuscript, usually the version that the author intends to submit to a journal. Thus, this draft is not peer-reviewed. For instance, Peer J defines pre-prints as follows:

A PeerJ 'PrePrint' is a draft that has not yet been peer reviewed for formal publication. Similar to preprint servers that already exist (for example arXiv.org), authors can submit draft, incomplete, or final versions of articles they are working on. preprints are not "publish ahead of print" articles, or articles that have been accepted and shown online before it has gone through typesetting, etc.

The term "post-print" is often used to refer to the revised version of the article, after the suggestions based on the peer reviewer comments have been included. In most cases, this is the version that has been accepted for publication. Thus, the post-print is very similar to the published article, with the published article including minor changes incorporated during copy editing.

Most publishers prefer to have pre-prints rather than post-prints on a public repository. Once the author has signed the copyright transfer document, he or she gives away a lot of rights to the publisher. As far as I know, copyright extends to the accepted version of the paper. Therefore, I think it is best to put up only pre-prints that have not been peer reviewed on public repositories. If you are planning to put up the revised version, that is, the version that has been accepted, it would be better to write to the journal and clarify before doing so.