← Analysis Reports

Page generated: 2026-08-27
Data (model runs) generated: 2026-07-23
Scoring inputs: ground truth 36c390d2, evaluate.py 5c52b890, rules 355f2e77

Page 39 — Interactive Comparison

AVS IS 758-781 1770 08.jpg · GT: 10 travelers · Best run per (model, phase) chosen by F1.

Claude
GPT
Gemini 2.5
Gemini 3.1
Claude Thinking
Legend — scored fields: match (char acc ≥ 0.999)   partial (0.7–0.999)   poor (< 0.7)   unscorable — GT blank, nothing to grade against
Name column (the row-matching key, not a scored field): exact (≥ 0.999)   close (0.85–0.999)   weak (0.70–0.85) — row still counted as matched for F1. A name below 0.70 is not matched at all; that row appears under unmatched extra row instead.