Agents & Tool Use
tau-Voice
tau-Voice tests whether voice agents can handle live customer service calls in retail, airline and telecom settings, where a call passes only if the database ends in the required state and the agent conveys the required information.
278items
6subjects
MITlicense
agents_and_tool_usedomain
audiomodality
textmodality
item-level responses released
Saturation status: No
Response matrix
Rasch analysis p = σ(θ − z + c)
8,382 responses, 80/20 split over cells · 6 subjects · 278 items · 12 conditions
AUC train
0.866
AUC test
0.851
Loading session strips…
Correct (1)Incorrect (0)Unobserved
Scale: 1 = correct · 0 = incorrect