MojiDiff

← experiments

openmoji-curation-v4-f9fc506-73aa9901

completed —

Hypothesis

empirical tail nomination plus contact-sheet review preserves legitimate outliers

Visual output

hist-alpha_fraction_18
hist-alpha_fraction_18.png
hist-alpha_fraction_72
hist-alpha_fraction_72.png
hist-bbox_area_72
hist-bbox_area_72.png
hist-bbox_height_72
hist-bbox_height_72.png
hist-bbox_width_72
hist-bbox_width_72.png
hist-component_count_72
hist-component_count_72.png
hist-ink_fraction_72
hist-ink_fraction_72.png
hist-path_count
hist-path_count.png
hist-segment_count
hist-segment_count.png
hist-total_path_length_estimate
hist-total_path_length_estimate.png

Written result

All 104 metric-tail candidates were visually confirmed legitimate and restored with reason-coded overrides. The reviewed primary manifest has 4,006 leakage-resistant rows.

State transitions

  1. completed2026-08-23T14:53:35.882649Zcompleted_reviewed_manifest

Run record

Verbatim from runs/openmoji-curation-v4-f9fc506-73aa9901/run.yaml, the record committed before launch.

run_id
openmoji-curation-v4-f9fc506-73aa9901
state
completed
hypothesis
empirical tail nomination plus contact-sheet review preserves legitimate outliers
parent_run
openmoji-audit-v2-f9fc506-3bc8e542
git_commit
None
code_identity_status
exact dirty pre-run snapshot was not recorded
config
configs/data/openmoji-curation-v4-reviewed.yaml
config_sha256
73aa9901131874d4…
source_revision
f9fc506a3f913be9897ab0181d611d4c910a4104
worker
gtc-local-cpu
resource_cap
local reversible curation only
artifact_policy
compact manifests versioned; renders persistent-local-only