you cannot breach the copyright, if so then what is the fun of copyright and journal policy, so better to be bound to the copyright rather to be cunning,
To the best of my knowledge, all journals make provision for the 'version' of the manuscript that can be shared freely. It could be the 'uncorrected proof', the 'accepted manuscript' e.t.c.I think it is always better to abide by the rules of the journal.
For non open access journal the author of accepted manuscript will sign a copyright transfer form. so, you cannot share full paper but you can share abstract and have discussion on the manuscript.
@ Pavan Kumar some publishers allow you self-archive certain versions of your accepted manuscript afer about a year of publication, while others would insist that in addition to self arhiving in places like RG, you also include a statement referring readers to their website, where the full versions are archived
Navigating through the discussions that took place with this question my insight says, since the authors are abided by the copyright rules of a specific non-open access journal where the article has been published, only summary or abstract or partial dataset may serve the purpose in a public platform like RG. It is always better share the published paper by one to one basis or hyperlinking to other full length article archival websites.
It depends on the rule, which is determined by the journal, for example which is given by Elsevier. Basically, it depends on what kind of access right you have to the journal. They said that it is commonly not allowed to share the paper publicly. Paper sharing may be okay through personal messages, based on a certain access right status.