metadata: migrate score_type -> score_pipeline (atomic stage contract; see mm-eval scorer docs/en/SCORING.md)

format v3 fix: remove stale pre-v3 config 'zeroshot' (1 parquet files) — it shadowed the v3 'default' config (README marked it default:true). Data now serves the flat v3 layout from data/.

Add manifest (metadata.json)

Upload dataset

metadata: 2026-07-07 revised taxonomy (task_type format vocabulary multiple_choice/yes_no/short_answer/free_form; score_type incl. none; re-verified protocol evidence; repaired prompt provenance)

metadata: add score_type/score_params/score_protocol (official grading protocol taxonomy); task_type is now format-only (judge_scored migrated to vqa + llm_judge). See dataset-upload skill references/metadata-json.md.

Update metadata.json

Push zeroshot config

initial commit