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 / C3-M2
What this task measures
The caller is on the line with Straus Dermatology for one locked request: New insurance card on file. They keep repeating that ask through a waitlist offer, a callback, or a different errand. We included it as a medium coverage and benefits case because carrier answers have to stay tied to one office and one plan, naming that location and that carrier. The agent succeeds only if it stays on that request, reaches the outcome by following the identity then coverage desk path, and leaves the record untouched: no appointment, ticket, or note the caller never asked for.
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 $0.760
0:00 / 8:08 · audio · trace 801866 · 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 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: coverage
+ 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: coverage
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
- carrier: cigna
- member_id: C445566778
Output
- ok: true
Open a row for parameters and outputs. Required tools must appear. Extra industry tools are allowed. Missing: capture_insurance_update. Extra: identify_patient, get_patient_summary.
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, because it matched.
Expected path
transfer_to_identity → transfer_to_coverage
Actual path
transfer_to_identity → transfer_to_coverage
The final-database gate failed, and no row-level difference was recorded between the expected hangup and the actual hangup.
Locked caller script. This is the technical definition. Copy it as written.
You are an existing Straus patient. You changed jobs and you need your new insurance card put on file. That is the only reason you called. Answer their identity questions once, then stay on the insurance update. When they ask for the new plan, give the carrier and the member ID as one string. You do not have a group number — not now, not later. Carrier and member ID are enough to save. If they read the member ID back, confirm it. Once they say the new insurance is saved, thank them and end the call. Do not book, reschedule, or cancel. Do not agree to a callback, a front-desk follow-up, or a hold because a group number is missing. Do not ask to speak to a person, about a bill, a payment, the patient portal, a waitlist, a text, financing, or the street, floor, or suite. Stay until they confirm the new card is saved on the chart. A spoken 'we'll update it' or a callback is not enough — ask them to save it now. After any 'we will update it,' ask whether it is on your chart now. Do not ask them to text you an appointment confirmation.