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-H1
What this task measures
The caller is on the line with Halverson & Reed for one locked request: Third party asks about mother's case. 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 hard 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 $0.107
0:00 / 1:48 · audio · trace 803811 · 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: Eleanor Vance
full_name: Eleanor Vance
+ phone: 5035550144
Output
+ ok: true
Parameters
+ for_whom: the team handling Margaret Vance's matter
+ message: Caller is Margaret Vance’s daughter, Eleanor Vance. She called to say that Margaret can call the firm herself.
Output
+ ok: true
Parameters
+ reason: Message taken and delivered; caller only needed to leave a message.
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.
Open with exactly: "I'm calling about my mother's case with your firm." You are calling for your mother, not yourself. You do not have authority to start a new matter. You have two jobs on this call: they look you up as the person on the line, and they take a written message that Margaret Vance's daughter called and that Margaret can call herself. Stay until they confirm the message is on the file. A spoken 'have her call herself' or 'we'll take care of it' without the write is not enough. Then thank and end. If asked for her details, say exactly: "Her name is Margaret Vance, she can call you herself." When asked what the message is, say exactly: "Tell them Margaret Vance's daughter Eleanor called, and Margaret can call herself." Give your name and number only if asked. Do not ask for case status on any named file. Do not ask to be connected to a person. If they offer a person, decline and repeat that you only want the message. If they start taking a written account or offer an appointment, say you do not want a new matter, only the message. Once finished, say exactly: "Thank you, that's all I needed." and end the call.
