Benchmark 100 paired web-search intents with measured Oracle token and tool costs

This commit is contained in:
Marcos Lohmann
2026-10-15 11:08:48 -03:00
parent 7f2ce6750b
commit 62258836bb
14 changed files with 18294 additions and 0 deletions

5
.gitignore vendored
View File

@@ -14,3 +14,8 @@ evidence/*
!evidence/RELATORIO.md
!evidence/validation-summary.json
!evidence/grok-websearch-models.json
# Resultados brutos do comparativo permanecem locais
benchmarks/runs/
benchmarks/intents-source.txt
benchmarks/plan.json