📑 Policy Reforms Toolkit

📅 Weekly Availability Mark when you are free each week and read off the best slots for the recurring call

📄 Document Viewer Browse documents with their extracted or annotated policies 📝 Prompt Viewer View and diff prompt versions side by side 🪞 Duplicate Review Triage suspected duplicate documents and pick the copy to keep 🌍 Geographic Rules Trace how a raw folder name becomes the countries sent to the prompt

⚖️ Extraction Comparison Compare model predictions against reference annotations ✏️ Annotation Review and annotate model-extracted policies 🔍 Annotation Comparison Compare both annotators' reviews of a document side by side

🔗 Consolidation Before/after view of Step-2 consolidated policy clusters 🧬 Consolidation Sweep Explore canonical policies, their timelines, mentions and families 📊 Production Progress Live country and mention completion, endpoint throughput and estimated time remaining 2b Step 2b Progress Post-consolidation stages, active countries and endpoints, logged calls and cost 🧩 Manual Consolidation Merge a sample of Denmark policies by hand and resolve fields 🧪 Consolidation Evaluation Audit a sweep's discards, entities and pairs, and the Step-2b repair 🆚 Variant Review Judge blind the pairs where the v008-retry and v007-repair sweeps disagree

⚖️ Importance Review Grade policy importance blind, then compare with the models, on the five-level scale (v002, A–E) or the four-level scale (v004, A–D) 🧑‍⚖️ Divergence Adjudication Decide who is right on each luna-vs-sol enrichment disagreement 🧭 Dimension Review Approve, edit or reject the proposed direction dimensions before freezing 🎯 Direction Review Grade policy direction scores blind, then compare with the scoring models 🕵️ Leakage Review Judge where each Step-3a date or figure absent from the input comes from