feat: add SWE-bench and TAU-bench benchmark suite#38

Open

Gui-Yue wants to merge 3 commits intoshareAI-lab:mainfrom

Gui-Yue:add_benchmark_test

Commits on Feb 13, 2026

feat: add SWE-bench and TAU-bench benchmark suite, fix OpenAI baseUrl version path matching
Gui-Yue
committed

Commits on Feb 26, 2026

refactor(benchmark): simplify to SWE-Verified + TB2 official flow, fix TB2 scoring, and trim dead TAU
Gui-Yue
committed

Commits on Feb 28, 2026

feat(benchmark): restore official TAU2, add token accounting for TAU/TB2, and gate
Gui-Yue
committed