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-E3
What this task measures
The caller is on the line with Straus Dermatology for one locked request: Card on file to hold an appointment. They keep repeating that ask through a waitlist offer, a callback, or a different errand. We included it as an easy 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 staying on reception, 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 $0.048
0:00 / 0:47 · audio · trace 801956 · click a span to play, drag to scrub, scroll to pan
Audio
Speech Audio
Agent Audio
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
+ 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: 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
stay on reception
Actual path
stay on reception
The actual hangup matches the expected hangup.
Locked caller script. This is the technical definition. Copy it as written.
Ask whether you need a credit card on file to hold an appointment. You are not reading a card number out on this call and you are not booking. Once you have the answer, thank them and end the call. Do not ask to book, reschedule, or cancel. Do not ask to add insurance, for a callback, to speak to a person, a bill, a payment, the patient portal, a waitlist, a text, financing, or the street, floor, or suite. Do not ask them to text you an appointment confirmation.
