List your company

Agent Task Success on tau2-bench Retail Domain

61.4Total Pass RateSkillsInjector
39.66445.30750.9556.593May 28, 2026

Evaluation results

MethodLinks
SkillsInjector2026.0561.4———
Graph of Skills2026.0560———
SkillRouter2026.0559.8———
Dense Cosine2026.0555.3———
LLM-as-selector2026.0555.1———
BM252026.0554.9———
Random-skill2026.0553———
No-skill2026.0551.2———
Full-library2026.0540.5———