Proposal: a clinical/medical discipline module (EQUATOR gates + medical disclosure venues + Chinese-literature resolver) #538
ktao732084-arch
started this conversation in
Ideas
Replies: 1 comment
|
Thank you for the proposal, and for your kind recognition of ARS. My suggestion would be to take the sibling route. The main consideration is that features for a specialized discipline are best maintained and judged by experts in that field, while ARS intends to remain on its general-purpose path. I would be glad to list your project in If, as you build and maintain the module, you find areas where the ARS core could be strengthened, you are always welcome to open an issue or ask questions here. Looking forward to seeing what you build. |
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
Thanks for building ARS — we read the code, not just the README (the fail-closed
verification_gate, thegold/↔heldout/eval split, thecompliance_agentPRISMA-trAIce/RAISE framework). It's the most engineering-mature project of its kind we found.We're a small non-commercial effort building AI-assisted-research support for clinical medicine (our end users are hospital clinicians, mostly writing in Chinese). Rather than rebuild a pipeline that already exists and is better than ours would be, we'd like to explore contributing a clinical discipline module upstream — prompted by CONTRIBUTING's "discipline-specific modules are encouraged."
Pieces that seem to fit your existing architecture:
compliance_modebranches alongside your PRISMA-trAIce path.nature_policy.md.unresolvablenarrowing).Per your POSITIONING "recorded non-goals", we'd keep anything institution/venue-specific (a hospital's promotion criteria, a journal's numeric AIGC threshold, predatory-journal lists) out of tree as user-supplied profiles — not in a PR.
Two questions: (1) would you want a medical module upstream, or as a sibling distribution? (2) if upstream, any preferred sequencing to keep PRs reviewable? We're fine with CC BY-NC and happy to meet the tests / eval / CI / CHANGELOG bar and start small. Thanks!
All reactions