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-M3
What this task measures
The caller is on the line with Straus Dermatology for one locked request: A bill the caller is sure they received. They keep repeating that ask through a waitlist offer, a callback, or a different errand. We included it as a medium 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 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 $0.325
0:00 / 3:48 · audio · trace 801991 · 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
Pass
All three gates matched, so the conversation counts as a pass.
Expected versus actual tool calls on this conversation.
Expected
Actual
Parameters
+ from: reception
- next_intent: billing
+ to: identity
Output
None recorded.
None recorded.
Parameters
None recorded.
None recorded.
Output
+ ok: true
Parameters
+ from: identity
+ to: billing
Output
None recorded.
None recorded.
Parameters
dob: 1985-11-03
dob: 1985-11-03
full_name: Sam Nguyen
full_name: Sam Nguyen
Output
- data: {"verified":true}ok: true
ok: true
Parameters
None recorded.
None recorded.
Output
- data: {"balance_cents":0}ok: true
ok: true
Parameters
None recorded.
None recorded.
Output
+ ok: true
Parameters
+ reason: caller_done
Output
None recorded.
None recorded.
Open a row for parameters and outputs. Required tools must appear. Extra industry tools are allowed. Extra: get_patient_summary, get_account_balance, end_call.
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_billing
Actual path
transfer_to_identity → transfer_to_billing
The actual hangup matches the expected hangup.
Locked caller script. This is the technical definition. Copy it as written.
You are an existing Straus patient and you are sure you were sent a bill — you remember opening a statement. Answer their identity questions once with your name and date of birth. If they ask for a second identity factor, give your ZIP code 11201, and confirm each readback. Then go straight to the bill. If they say the balance is zero, insist once that you definitely received something and ask them to look again. After they look again and confirm the balance is still zero, accept that, thank them, and end the call. If they offer a callback, a payment link, a transfer to a person, or to book, move, or cancel an appointment, decline and stay on the bill. Do not volunteer a callback, do not give a callback number, and do not ask for a text, the street, the floor, the suite, an office id, a waitlist, or financing. Stay on the bill. If they try to book, transfer, or wrap up before they have looked at the account again, bring them back to the statement. Do not ask them to text you an appointment confirmation.