-
register part2 configs
-
annotate_rafael_part2: 240 more passages (additive, part1 untouched)
-
annotate_aivin_part2: 240 more passages (additive, part1 untouched)
-
annotate_aj_part2: 240 more passages (additive, part1 untouched)
-
annotate_paddy_part2: 240 more passages (additive, part1 untouched)
-
drop extractor-duplicate spans (same offsets)
-
drop extractor-duplicate spans (same offsets)
-
drop extractor-duplicate spans (same offsets)
-
drop extractor-duplicate spans (same offsets)
-
drop extractor-duplicate spans (same offsets)
-
annotate sets: mutually exclusive, stratified
-
annotate sets: mutually exclusive, stratified
-
annotate sets: mutually exclusive, stratified
-
annotate sets: mutually exclusive, stratified
-
annotate sets: reason-covered resample
-
annotate sets: reason-covered resample
-
annotate sets: reason-covered resample
-
annotate sets: reason-covered resample
-
spans: add luna_reason (+usage/drop/specificity backfill)
-
spans: add luna_reason (+usage/drop/specificity backfill)
-
spans: add luna_reason (+usage/drop/specificity backfill)
-
spans: add luna_reason (+usage/drop/specificity backfill)
-
spans: add luna_reason (+usage/drop/specificity backfill)
-
spans: add luna_reason (+usage/drop/specificity backfill)
-
register per-annotator configs
-
annotate_rafael: 240 stratified passages (24 shared overlap)
-
annotate_aivin: 240 stratified passages (24 shared overlap)
-
annotate_aj: 240 stratified passages (24 shared overlap)
-
annotate_paddy: 240 stratified passages (24 shared overlap)
-
to_annotate: single holdout_annotate split
-
to_annotate: remove split files (now single holdout_annotate)
-
to_annotate: remove split files (now single holdout_annotate)
-
to_annotate: remove split files (now single holdout_annotate)
-
to_annotate: restructure to annotation shape (text only, probe_tag bands)
-
register to_annotate config
-
to_annotate holdout: probe-v3 Luna-verdict spans with raw passages
-
to_annotate val: probe-v3 Luna-verdict spans with raw passages
-
to_annotate train: probe-v3 Luna-verdict spans with raw passages
-
README: document text field
-
add text (joined passage) alongside tokenized_text
-
README: document human_verdict/human_note
-
spans: add human_verdict + human_note (jdc_verdicts, prompt_probe_sync_verdicts)
-
human473 gliner holdout (annotator190+jdc283)
-
initial commit