Working paper / 2026
Multi-Industry Voice Agent Simulation Bench
Abstract
Multi-Industry Voice Agent Simulation Bench (MIVAS) is an indicator of voice AI performance across economic sectors. MIVAS is comprised of tasks and verifiers with granular rewards executed in production-grade multi-agent RL environments across industries where voice AI is being adopted the fastest. Each task places a model harness inside a multi-agent graph with production-length prompts, tools, and isolated state. Deterministic verifiers score tool use, handoffs, and final database state; pass@1 and passk measure single-run capability and repeated-run reliability.
Clone the repository, pair a model harness with an industry task suite, and reproduce the evaluation environment.
Healthcare / C5-H2
What this task measures
The caller is on the line with Straus Dermatology for one locked request: Balance question then move tomorrow's visit. They keep repeating that ask through a waitlist offer, a callback, or a different errand. We included it as a hard billing and payments case because balances and payment links have to come from the chart, at the amount on file. The agent succeeds only if it stays on that request, reaches the outcome by following the identity then billing then scheduling path, and leaves the record exactly as this case requires: the right row written or cleared, and nothing extra.
One model harness at a time, with a representative pass and fail conversation when both exist. OpenAI Realtime 2.1 is the default view.
Recording
Conversation cost $0.388
0:00 / 3:27 · audio · trace 801966 · click a span to play, drag to scrub, scroll to pan
Audio
Speech Audio
Agent Audio
Handoffs
Tool Calls
Errors
Transcript
How a conversation passes
Combined pass is the logical and of three deterministic gates. Tools, handoff path, and final database state must all pass on the same conversation. One miss fails the call.
Combined
Fail
This sample fails because tools and handoff path and final database state did not match. Combined cannot pass unless every gate does.
Expected versus actual tool calls on this conversation.
Expected
Actual
Parameters
+ from: reception
- next_intent: billing
+ to: identity
Output
None recorded.
None recorded.
Parameters
+ dob: 1990-04-12
+ first_name: Jordan
+ last_name: Lee
+ zip: 10016
Output
+ ok: true
Parameters
None recorded.
None recorded.
Output
+ ok: true
Parameters
+ from: identity
+ to: billing
Output
None recorded.
None recorded.
Parameters
None recorded.
None recorded.
Output
None recorded.
None recorded.
Parameters
dob: 1990-04-12
dob: 1990-04-12
full_name: Jordan Lee
full_name: Jordan Lee
+ second_factor: 10016
Output
- data: {"verified":true}ok: true
ok: true
Parameters
None recorded.
None recorded.
Output
- data: {"balance_cents":12500}ok: true
ok: true
Parameters
- line_item_id: li_visit
+ line_item_id: li_noshow
Output
ok: true
ok: true
Parameters
- location_ids: ["loc_park_ave"]
Output
None recorded.
None recorded.
Parameters
- appointment_id: 1
- new_end: 2026-08-24T09:30
- new_start: 2026-08-24T09:00
Output
- data: {"status":"rescheduled"}- ok: true
Parameters
+ line_item_id: li_visit
Output
+ ok: true
Parameters
+ line_item_id: li_visit
Output
+ ok: true
Open a row for parameters and outputs. Required tools must appear. Extra industry tools are allowed. Parameter miss: explain_charge. Missing: transfer_to_scheduling, find_slots, reschedule_appointment. Extra: identify_patient, get_patient_summary, explain_charge, explain_charge.
Full industry architecture, same graph as the industry page. The left diagram marks the expected specialist path in green. The right diagram marks the path this conversation actually took in green through the shared prefix, then red from the first wrong hop.
Expected path
transfer_to_identity → transfer_to_billing → transfer_to_scheduling
Actual path
transfer_to_identity → transfer_to_billing
Rows that differ between the expected hangup and the actual hangup.
appointments · 1
Expected
Actual
- end: 2026-08-24T09:30
+ end: 2026-08-20T10:20
- start: 2026-08-24T09:00
+ start: 2026-08-20T10:00
Locked caller script. This is the technical definition. Copy it as written.
You are an existing Straus patient. Start by asking what you owe — the total amount and a short explanation of the charges. Answer their identity questions once with your name and date of birth, confirm if they read it back, then go straight to the bill. As soon as you hear the amount and any short explanation, immediately say you also cannot make August 20 appointment at Park Avenue and you need to move it to a later date, not cancel it. Do not wait for them to finish offering payment, financing, a dispute, or a text. Do not agree to pay, dispute a fee, or take a payment link. Take the first appointment time they offer. Do not ask for a different day, time, or office, and do not suggest any date of your own. Confirm that first time out loud, then thank them and end the call. Do not ask to cancel. Do not ask for a callback, to speak to a person, to change insurance, about the patient portal, a waitlist, or for a text. Do not ask for the street, floor, suite, or office id. Do not ask to move the visit until they have named the charge. After the move, stay until the new time is on the calendar. A listed opening is not a booking. Do not thank them, say goodbye, or hang up until they confirm a time is on the calendar. If they only list openings, say book the first one. Do not ask them to text you an appointment confirmation.