Seeded Dataset

LeWiDi Moderation

Content moderation · imported 2026-05-02 · sha256:1c83…77ee

← DatasetsSeeded Demo
Records
412,330
Actors
1,842
Outcomes
Annotator consensus
Imported
2026-05-02
Status
adapted

Schema Mapping

source → DecisionEvent
source fieldDecisionEvent field
annotator_idactor_id
labeldecision
consensus_labelground_truth
consensus_nconsensus_n
confidencecontext_features.self_confidence

Data Quality

Actor Coverage94.0%
Outcome Coverage71.0%
Temporal Coverage88.0%
Ground Truth Confidence72.0%

Calibration Readiness

78
/ 100
Ready w/ caveats

Annotators have stable IDs and repeated assignments. Truth is consensus-derived, which introduces a confidence ceiling on ECE estimates.

Blocking Issues
  • · Consensus-based truth only; no external outcome resolution

Sample Records

first 3 raw rows
item_idannotator_idlabelconfidenceconsensus_labelconsensus_n
i_4421r_127unsafe0.82borderline7
i_4421r_044borderline0.61borderline7
i_4422r_127safe0.7safe5