Skip to content

Add reviewer calibration assistant#574

Open
catcherintheroad-hub wants to merge 1 commit into
SCIBASE-AI:mainfrom
catcherintheroad-hub:catcherintheroad/reviewer-calibration-assistant
Open

Add reviewer calibration assistant#574
catcherintheroad-hub wants to merge 1 commit into
SCIBASE-AI:mainfrom
catcherintheroad-hub:catcherintheroad/reviewer-calibration-assistant

Conversation

@catcherintheroad-hub
Copy link
Copy Markdown

Summary

/claim #16

Adds a distinct reviewer calibration and inter-rater agreement assistant for the AI-Powered Research Assistant Suite. The module validates synthetic peer-review packets before AI-generated review summaries, reputation updates, or author-facing recommendations are released.

This slice focuses on reviewer calibration rather than another broad research assistant, evidence binder, structured-abstract checker, model-assumption diagnostic, dependency guard, external-validity transfer assistant, research-gap finder, or reproducibility executor.

What it checks

  • pairwise reviewer agreement across clarity, rigor, novelty, reproducibility, and recommendation
  • recommendation consensus ratio
  • severe score spread by scoring dimension
  • domain-template drift
  • evidence-anchor coverage for reviewer claim checks
  • duplicate reviewer records
  • whether disputed packets should remain double-blind during calibration

Reviewer artifacts

  • reviewer-calibration-assistant/reports/reviewer-calibration-report.json
  • reviewer-calibration-assistant/reports/reviewer-calibration-report.md
  • reviewer-calibration-assistant/reports/reviewer-calibration-summary.svg
  • reviewer-calibration-assistant/reports/reviewer-calibration-demo.mp4

Demo video raw URL:
https://raw.githubusercontent.com/catcherintheroad-hub/SCIBASE.AI/catcherintheroad/reviewer-calibration-assistant/reviewer-calibration-assistant/reports/reviewer-calibration-demo.mp4

Validation

  • npm test
  • npm run demo
  • node --check reviewer-calibration-assistant/src/calibration.js
  • node --check reviewer-calibration-assistant/demo/generate-demo.js
  • node --check reviewer-calibration-assistant/demo/write-svg.js
  • git diff --check
  • ffprobe verified the committed MP4 as H.264, 1280x1280, 8 seconds

Boundaries

Dependency-free Node implementation using synthetic data only. It does not call external AI APIs, ingest private manuscripts, access credentials, or integrate with payment systems.

@catcherintheroad-hub
Copy link
Copy Markdown
Author

Maintainer-ready validation refresh for #16 / PR #574.

Current external status checked now:

  • Algora claim page: https://algora.io/claims/JdW1FuCwZs1qe65s
  • Public claim text reports: Total prize pool $2,675, Total paid $0, Status Pending.
  • PR metadata remains open, non-draft, mergeable, and 🙋 Bounty claim labeled.

Fresh local validation on the current branch:

  • PATH=/opt/homebrew/opt/node@22/bin:$PATH npm test -> passed (reviewer calibration assistant tests passed)
  • PATH=/opt/homebrew/opt/node@22/bin:$PATH npm run demo -> passed and regenerated the calibration report
  • PATH=/opt/homebrew/opt/node@22/bin:$PATH node --check reviewer-calibration-assistant/src/calibration.js
  • PATH=/opt/homebrew/opt/node@22/bin:$PATH node --check reviewer-calibration-assistant/demo/generate-demo.js
  • PATH=/opt/homebrew/opt/node@22/bin:$PATH node --check reviewer-calibration-assistant/demo/write-svg.js
  • git diff --check

Demo evidence remains reachable:

Scope reminder: this is the reviewer calibration / inter-rater agreement slice for #16, focused on pairwise agreement, recommendation consensus, score divergence, domain-template drift, evidence-anchor coverage, duplicate reviewer records, and double-blind escalation for disputed review packets. It is intentionally separate from the other #16 assistant slices already submitted.

Could a maintainer please review PR #574 for merge/reward decision, or point me to the exact blocker if one remains? I can turn around any requested revision quickly.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant