# Cumulative experiment state S002

Evidence status: illustrative. Handoff produced by round 001 from S001.

- Candidate policy for the next batch: P002. Retain R01 and R04, amend R02 and R03, and add R05.
- Carry forward F01 (axis label promoted to title), F02 (statistical form promoted to variable), and F04 (unsupported counted entity) as failure modes to look for in fresh samples.
- Human adjudication H01 resolved F03 for this round. Preserve the visible cross-reference in D2 and leave the star thresholds unspecified. Gold and schema revisions remain as recorded in S001.
- Human spot check H02 accepted the selected C1 fields against the data snapshot. This check covers those fields only.
- Rejected approaches: substituting a generic title for any data snapshot without a visible title and supplying a counted entity from bar labels alone. Both would add unsupported content.
- Open challenge from F04: test R05 on figures with explicit units as well as unlabeled axes, and on tables with row-specific units. The counted entity for C2 remains unspecified.
- Trace back to the round 001 findings, deliberation, judge reviews, and extractions under P001. P002 awaits a subsequent round. An independent evaluation remains pending.
