code wiki / _hdl_build / nx_peer_review.nx

nx_peer_review.nx

buildroot/runtime/_hdl_build/nx_peer_review.nx

2873 B44 linesdepth 3pulls 5 transitivereach 4 importersview sourcekind librarytopic peer
docsdependenciesstructsconstsfunctions

about

nx_peer_review.nx -- who watches the watchers. The evaluators must themselves be checked, or the honest-S-class verdict is just one organ's opinion (operator). Three mechanisms: DUAL GRADE -- the Examiner AND Claude grade the same thing; AGREE = trustworthy, DIVERGE = a tie-break flag (Claude or the operator breaks it). Divergences map the Examiner's growth: a divergence usually means the Examiner LACKED a SOTA baseline (the RESEARCHER's job to supply) or the EVIDENCE/citations (the LIBRARIAN's job) -- so the Examiner is built up IN CONJUNCTION with the Researcher + Librarian. MUTUAL JUDGE -- the CRITIC judges the Examiner's grading (is it triangulated + red-teamed + counter-free = sound?), and the Examiner grades the Critic (does it clear the teammate bar?). Checks & balances on the judges themselves. TIE-BREAK -- when two graders disagree, a human/Critic verdict resolves it; agreement stands. license_tier: ORIGINAL Refs: separation of powers; peer review.

dependencies 3 imports · 2 importers

nx_sclass_grader.nx nx_critic.nx nx_team_scorecard.nx nx_peer_review.nx nx_paper_panel.nx nx_publish_review_test.nx

imports: nx_sclass_grader.nxnx_critic.nxnx_team_scorecard.nx

imported by: nx_paper_panel.nxnx_publish_review_test.nx

structs

none

consts

18const PR_AGREE: i64 = 0
19const PR_DIVERGE: i64 = 1
21const PR_OK: i64 = 0
22const PR_NEED_SOTA: i64 = 1 // RESEARCHER must supply the SOTA baseline
23const PR_NEED_EVIDENCE: i64 = 2 // LIBRARIAN must supply the evidence/citations

functions

25func pr_agreement(grade_a: i64, grade_b: i64) -> i64 { if grade_a == grade_b { return PR_AGREE } return PR_DIVERGE }
called by 2: pp_verdictmain
28func pr_resolve(grade_a: i64, grade_b: i64, tiebreak: i64) -> i64 { if grade_a == grade_b { return grade_a } return tiebreak }
called by 2: pp_verdictmain
31func pr_divergence_cause(has_sota_baseline: i64, has_evidence: i64) -> i64
39func pr_critic_judges_examiner(grading_reproductions: i64, grading_counter_strength: i64, grading_redteamed: i64) -> i64
called by 1: main calls 1: crit_status2
44func pr_examiner_grades_critic(critic_score: i64) -> i64 { return tsc_passes(critic_score, TSC_BENCHMARK) }
called by 1: main calls 1: tsc_passes