Repository navigation
Integrate GENIUS with persistent participant sessions and library discovery - #70
Merged
Merged
Conversation
Preserve contributor history; production merge remains subject to review.
Preserve the contributed 23-stage protocol, isolate feedback contexts, account for model-call budgets, and retain diagnostic limitations. Add offline acceptance and explicit unavailable-statistics results. Co-authored-by: Codex <noreply@openai.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Integrate the GENIUS contribution from #26 into the existing study library and hosted runner. Preserve the original contributor commits and attribution; use the canonical library ID study_016 so web and desktop catalogs discover it.
The hosted runner now selects the study's explicit persistent protocol instead of flattening it into independent trials. It retains all 23 participant administrations, isolated feedback contexts, generated baseline scoring, per-stage progress and participant checkpoints. Budget checks count up to 28 model calls per participant. Diagnostic limitations and unavailable-statistics status remain visible in analysis.
This contribution includes originating-researcher-authorized pseudonymous human reference extracts and a text-derived scorer. These are evaluator assets, not participant inputs. The project owner explicitly approved publishing these assets with the integration on November 5, 2026. This approval and the original rights statement are not a new privacy certification.
Validation: 208 offline tests pass; actual playground CLI --simulate completes both 23-stage conditions, returns not_scored for developer fixtures and leaves replicationRate null. Mock-provider tests verify 46 participant calls plus 5 fresh feedback calls and OpenRouter routing. Independent GPT-6.1 Terra review findings were addressed.
Not established: live HTTP provider behavior, full 801-participant R/MDMB evaluation, historical equivalence, scorer validity on generated essays or against human ratings. No paid model run or installer release performed.
Generated with Codex.