Nov 14 3 days ago 03:40 analysis/: trace-behaviour scripts, outputs and T10 route notes 03:40 README: behaviour section (checking after drafting, T10 routes, query presupposition) 03:09 README: citation precision definition 03:08 README: framed as three models in Claude Code with hedwigai workbooks; citation precision over fetched citations 03:02 README: time and tokens label 03:02 README: post-training framing (Particle-1 vs its DeepSeek base, GLM as reference), accuracy as % of rows right, workbook environment, MCP access, setup notes trimmed Nov 13 4 days ago 18:28 hedwigai harness-model benchmark v2: 180 graded units, 195 runs, docs, traces, keys, grades 18:28 initial commit