τ-bench
Sierrafrom Sierra
ττ-benchNew↗Agent-building agents
GitHubSubmit Results
NEW
ττAgents building agents →τ-i18nVoice agents across six languages →τ-EntityCapturing entities over voice →

Community Spotlight

Projects from the research community that build on τ-bench and extend agent evaluation to new domains and settings.

Research · October 2026Read more →

SEATauBench

Extends τ²-bench to five target languages across Airline, Retail, and Telecom. Four localization settings separate dialogue, tool understanding, and domain reasoning. Published at EMNLP 2026.

My Chiffon NguyenAulia AdilaSaksorn RuangtanusakKittiphat LeesombatwathanaVissuta Gunawan LimPatomporn PayoungkhamdeeSamuel Cahyawijaya
My Chiffon Nguyen, Aulia Adila, Saksorn Ruangtanusak, Kittiphat Leesombatwathana, Vissuta Gunawan Lim, Patomporn Payoungkhamdee & Samuel CahyawijayaSEACrowd · AI Singapore · SCB DataX, SCBX Group · Chulalongkorn University · VISTEC · University of Indonesia · Cohere
Research · September 2026Read more →

τ-Rec

Extends τ-bench to conversational recommendation, evaluating whether agents can elicit user preferences, use catalog tools, follow policy, and make verifiably correct recommendations across multi-turn interactions. Published at ACM RecSys 2026.

Bharath Sivaram NarasimhanKarthik Narasimhan
Bharath Sivaram Narasimhan & Karthik NarasimhanIndependent Researcher · Princeton University

For questions or feedback, contact [email protected]