01 / LIVE CASE STUDY
A family trip,
remembered.
A family and their travel contacts return across eight conversations. The questions below are exact benchmark samples, with reference answers revealed on demand. Browse all QA and dialogue ↗
VoxPolyBench dataset on Hugging Face ↗
Listen across sessions
Listen to a full session or open its individual turns. Each turn has a separate audio player and speaker name; dialogue transcripts are not shown here.
Loading session and turn audio…
Listen to 10 personalized spoken questions
The demo audio is synthesized speech. The complete benchmark contains 18 cases; only this case is selected for public playback.
Questions grounded in the trip
These items test who spoke to whom, what changed, and what each participant prefers.
SPEAKER ATTRIBUTIONWho asked about arranging a taxi van for the return trip?
Nora Patel
ADDRESSEE ATTRIBUTIONWho was Nora Patel speaking to when she said she’d really like a quiet room away from the elevator?
Tessa Kim
MULTI-HOPWhere can you stay to avoid elevator noise and what time is the walking tour that keeps you dry?
a quiet room away from the elevator; a 3 p.m. walking tour
MEMORY UPDATEWhat is the later plan after the earlier 11:30 entry slot?
a 3 p.m. walking tour
PERSONALIZATIONWhat kind of update did I say I prefer?
a written itinerary
Reference answers are hidden until opened. These questions are from the accepted family-trip case.