Oct 21 16 days ago 15:23 Rebuild data from the paper's 100 problems (fixes GLM stats, drops stray problem) Jul 12 01:05 Regenerate graphs after abc392_b→abc364_b swap 00:25 Regenerate graphs after abc392_b→abc364_b swap Jul 11 23:57 Regenerate graphs after abc392_b→abc364_b swap 21:02 Regenerate transition graphs: glm52 gap labels were stale (predated label files) Jul 8 20:43 Fix build: INCLUDE_MODELS filter compared absolute path, dropped all files Filter used Path(f["path"]).parts[0], but "path" holds an absolute
filesystem path so parts[0] is always "/" — no model ever matched,
so every trajectory was excluded and the space showed "No files found".
Use the relative "display" field instead. 16:54 build Jun 12 17:57 fix node text: state name bigger, counts as svg bg, larger special nodes Jun 11 01:45 sonnet46 full regen 00:42 exclude cpp 00:37 add sonnet46 graphs Jun 10 18:27 Sync graphs with M4 NLP fix; rename sonnet46 → sonnet46_no_thinking trajectories Aryarya +1: Claude Sonnet 4.6 Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> 14:59 Add gap label distributions to node detail panels; exclude sonnet from merged Aryarya +1: Claude Sonnet 4.6 Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> 14:11 Weighted label pct across models; fix efficiency token/solve-rate denominators 05:56 Regenerate graphs with per-label edge distribution table 05:38 Exclude cpp; med output tokens + CI; median+CI stats 05:31 Median + CI stats; remove sync_space_graphs.sh; rebuild from results_full 05:27 Fix: exclude ocaml_prev, deduplicate 0-turn timeout retries in stats 05:24 Sync latest data and graphs Jun 9 16:24 Rebuild from results_full: 1907 trajectories across 4 models × 4 languages 16:23 Restore graphs button + page; rebuild with full 100-problem results Jun 5 02:42 full 100-problem results + analysis artifacts, 100KB output truncation May 26 19:06 Add C++ trajectories (10 per model) May 22 18:54 Update Java trajectories to v4 runs 17:43 Add permalink support: URL hash syncs path+step, copy-link button 
