betaFeedback
DocsMachine contribution
docs / machine contribution

Machine contribution

At Valency, we believe in the potential of AI to help all areas of science and research. Many people are starting to use AI in their everyday workflows, and we firmly believe everyone wants to declare and attest to the contributions of AI to their work.

Every submission to the Hub says which AI models and systems contributed to the work and what they did, or says that none did. Here is how you declare it, what the Hub records, and how readers see it.

01Where you declare it

In the AI contributors step of the submission form. It needs nothing from the parser, so you can fill it in while your upload is still being parsed. Choose each model or tool once, select the contributor roles it helped with, and add it. Several models can share a role, and Edit roles changes a model’s selection later. To check the declaration one role at a time, open Review assignments by role. The declaration is required: the Hub refuses to finalize a submission that neither lists contributors nor affirms there were none.

02Picking from the catalog

The picker searches a catalog of models alongside bare entries for AI coding agents, AI IDEs, AI research assistants and LLM-agent frameworks. A bare entry records the tool without asserting a model version. Choosing an entry records its default components. A contributor already added to one row is offered first in every other row, so crediting it with several roles takes one pick per row.

If you used more than one version of a model, such as Claude Fable 5 and 5.1, choose the family’s multiple versions entry (for example, Claude Fable (multiple versions)) instead of listing each version. It names the model family without asserting a version.

If a search finds no match, choose Add “…” as a custom contributor and the name you typed is recorded as written. Plain software that is not AI, such as numerical libraries, notebooks and statistics languages, is not in the catalog: it is not an AI contributor.

03Saying what it did

Every contributor you list has at least one role, the row you added it in; removing it from its last row removes it. The form offers ten rows, by short label, each standing for a CRediT role: Concept (Conceptualization), Data (Data curation), Analysis (Formal Analysis), Investigation, Methods (Methodology), Software, Validation, Figures (Visualization), Drafting (Writing — original draft) and Editing (Writing — review & editing). Each role name shows its definition, taken from the NISO CRediT definitions.

04If no AI was used

Check None of my tools were AI. That answer is recorded on the version you submit. It is not itself shown on the work’s page, which says “none listed by the authors” when a work has no AI contributors.

05What is recorded, and how it is shown

Every contribution you declare is recorded as a collaborator, not an author, and is self-declared: the Hub does not verify it. Each AI contributor is recorded on the work itself rather than on one version (for a new version, once that version is published), so every published version shows the same list.

The work’s page lists its AI contributors with their CRediT roles, and its machine contribution statement marks each one as self-declared at submission.

06What is not built yet

Declarations are typed in by the submitter. Capturing them from an agent session’s traces, which needs the command-line submit flow, is not available, and neither is any verification beyond self-declaration. Both are planned; neither changes what you need to do today.