Hyper-τ-bench
Sierra / Coding · Builds and tests working customer-service agents from business materials.
What it measures and how
Measures agent development and debugging, separate from direct service execution in τ³-Banking. Tools, runtime, and acceptance environments are fixed; best results from different systems cannot be combined.
How this evidence is used
Under review: could add coding and business-system development evidence. Raw results, licenses, and fixed conditions require review first; new publication does not automatically trigger scoring.
Evaluation limits & attribution
Under review: could add coding and business-system development evidence. Raw results, licenses, and fixed conditions require review first; new publication does not automatically trigger scoring.
Data license: Leaderboard-data usage terms need confirmation.
Results are published by Sierra. Original scores and the NexGateHub consensus index use different scales and cannot be added directly.