Update README.md

Update README.md

Update README.md

update README.md

Add leaderboard and paper links; remove em dashes

Clarify that the runtime tool layer is not part of the release

Update README.md

Document Archipelago harness, run limits, and judge configuration

Update README.md

Update README.md

Add reasoning-effort labels to model names in the dev-set vs held-out score table

Update README.md

Trim card: drop Quick Links table, Expert Authorship, Quality Control, Example Record, World inventory, per-task breakdown, realism constraints, and eval/leaderboard sections; rename Task Taxonomy to Task Categories with dev-vs-held-out double-header table; move At a Glance, Task Structure and Splits up; add blog link

Update README.md

Update README.md

Update README.md

Add per-task dev-set breakdown (criteria, input files, task-specific files, est. hours) and blog link; remove empty Legal Disclaimer

Add the 16 per-task attachment files under task_files/; all context_files now resolve, drop the not-included caveat

Drop criterion depends_on and golden_response_prompt_addendum from task records and card schema

Add dev-set data: 10 task records (data/dev.jsonl + tasks/), 90 world files, LICENSE; card updated for real schema and the not-included input files

Revise card against latest paper: Loop Harness only, judge validation + model configs as own sections, dev-set vs held-out scores, corrected Mean Criteria@3 sampling and judge visibility, condensed QC/selection, placeholders removed

Update card from paper v6.1: 160-task/10-world held-out set, 10-task World 9 (Sterling Marsh LLP) dev set, new category taxonomy, 42-expert stats, judge/eval protocol, full leaderboard, harness-not-released policy

Add draft dataset card (README) + LFS attributes for jsonl/csv

initial commit