Verification check

feature · paper proposed

Declaration e66d89b51a90 has not been accepted by anyone. · no paper read under it

Does a re-run stand behind each claim, read beside its warrant?

A capability that revises existing artifacts. It runs once, and then it is part of the schema.

Part of Measures — What can be measured over the tree - its coverage, its wording, whether its code re-runs?

zmainen/claim-graphs#21 — the decision is taken there; the propagation is taken here.

How it works

For every claim it reads the reproduction records the claim carries and the verification provenance the audited run wrote (verification//provenance.json), and writes one verdict into check_verification:: reproduced (a record marked verified whose provenance shows the script ran and measured the value), partial (recorded verified but not observed running, or a run that executed without measuring the value), mismatch (a re-run that disagreed with the paper — the site shows it as contested-by-verification), blocked, unattempted, or unrecorded (no record at all). check_verification_from: lists the record statuses and provenance it read. It never touches warrant:.

Mechanical: no model and no prompt. The rule is precedence over the evidence a claim carries — the most adverse informative verdict winning — so a claim with both a blocked and a mismatch record reads mismatch. The dopamine paper's single dSTORM mismatch is the first real case. Export mapping — how a check reaches the MIRA and OXA encodings — follows.

Rests on

Feeds

Nothing — this is a leaf.

How it is defined

What this layer reads besides its dependencies. Each is a declared input: its content is hashed into every run, so editing one makes those runs stale.

extract/claim_graphs/verification_check.pythe script that runs itnot in the repository

The declaration names this path and the repository does not have it. An input that does not exist hashes to nothing, so it cannot make a run stale — the layer is declared to depend on something it is not in fact tracking.

What it produces25 checks

One paper, as the worked example — Ejdrup, v1. Read from runs/ejdrup-2026-dopamine/verification-check.output.json · 4 KB.

# slugcheck_verificationcheck_verification_from
1 d1r-tracks-da-50ms-delay partial ["record:verified"]
2 d2r-initialization-unjustified partial ["record:verified"]
3 d2r-insensitive-to-brief-pauses partial ["record:verified"]
4 d2r-integrates-over-seconds partial ["record:verified"]
5 d2r-occupancy-higher-in-vs blocked ["record:blocked"]
6 dat-clustering-greater-in-vs mismatch ["record:blocked","record:mismatch"]
7 dat-immunostaining-dorsoventral-gradient blocked ["record:blocked"]
8 dat-nanoclustering-slows-clearance partial ["record:blocked","record:partial"]
9 ds-lacks-pervasive-tonic-da partial ["record:blocked","record:verified","provenance:PASS"]
10 ds-vs-vmax-ratio-assumed partial ["record:verified"]
11 fscv-matches-may-wightman-1989 blocked ["record:blocked"]
12 hypothesis-d1-d2-temporal-distinction unrecorded []
13 hypothesis-nanoclustering-regulates-vmax unrecorded []
14 hypothesis-vmax-explains-regional-difference unrecorded []
15 interprets-cragg-rice-vmax-ratio unrecorded []
16 interprets-may-wightman-1989-fscv unrecorded []
17 low-burst-no-spillover-high-burst-does partial ["record:verified"]
18 nanoclustering-constant-vmax-constraint partial ["record:verified"]
19 nanoclustering-model-varicosity-scale partial ["record:verified"]
20 vmat2-gradient-absent blocked ["record:blocked"]
21 vmax-modulation-larger-impact-in-vs partial ["record:verified"]
22 vmax-only-parameter-driving-regional-difference reproduced ["record:verified","provenance:PASS measured"]
23 vs-low-active-fraction-resembles-ds-distribution blocked ["record:blocked"]
24 vs-lowest-percentiles-above-10nm partial ["record:blocked","record:verified"]
25 vs-maintains-pervasive-tonic-da partial ["record:blocked","record:verified","provenance:WARN"]

Across the corpus

8 not run · 1 stale · 1 blocked upstream·a paper links to its own cell, where this layer's output for it is rendered

Inputs and outputs

Reads, besides its dependencies
Produces
  • runs/{paper}/verification-check.output.json

One per paper — the table above links each one that exists.

Views
  • table — rendered above, over the 25 checks in the artifact

Running it

The command comes from the declaration, so this text and what actually runs cannot diverge. pipeline.py run also runs the unmet dependencies first.

python3 scripts/pipeline.py run <paper> verification-check

Underneath, that runs cd extract && python3 -m claim_graphs.cli verification-check --paper {paper}.