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.
Legal / C1-M3
What this task measures
The caller is on the line with Halverson & Reed for one locked request: Cut-off number then complete. They keep repeating that ask through a callback offer, a message left for someone else, or advice on a different matter. We included it as a medium reception and routing case because intake has to wait until the caller is classified; messages, wrong numbers, and represented parties are everyday reception work. The agent succeeds only if it stays on that request, reaches the outcome by staying on reception, 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 $0.163
0:00 / 2:46 · audio · trace 803841 · 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
full_name: Wendell Marsh
full_name: Wendell Marsh
+ phone: 2065550384
Output
+ ok: true
Parameters
+ for_whom: New cases team
+ message: Wendell Marsh called about a fall at a grocery store and wants a callback. Best callback number: 206-555-0384.
Output
+ ok: true
Open a row for parameters and outputs. Required tools must appear. Extra industry tools are allowed.
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.
Open with exactly: "I want to leave a message for whoever handles new cases." You called only to leave that message. You do not want a new matter, a screening, or an appointment. The first two times you are asked for your number, say exactly: "two oh six, five five five, zero three." The third time, say exactly: "Wendell Marsh, two oh six, five five five, zero three eight four." When asked what the message is, say exactly: "Tell them Wendell Marsh called about a fall at a grocery store and wants a call back." Do not volunteer who it was against, which state, or when it happened. If asked about a new matter, who it would be against, which state, when it happened, or offered an appointment or a person instead of the message, say exactly: "No, just the message please." Stay until they confirm the message is on the file. A spoken 'we'll take care of it' without the write is not enough. Once finished, say exactly: "Thank you, that's all I needed." and end the call.
