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-H4
What this task measures
The caller is on the line with Halverson & Reed for one locked request: Injured relative cannot speak. 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 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 $0.060
0:00 / 1:18 · audio · trace 803827 · 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
Fail
This sample fails because tools did not match. Combined cannot pass unless every gate does.
Expected versus actual tool calls on this conversation.
Expected
Actual
Parameters
- full_name: Corinne Halberg
Output
None recorded.
None recorded.
Parameters
+ reason_code: caller_request
Output
+ ok: true
Open a row for parameters and outputs. Required tools must appear. Extra industry tools are allowed. Missing: lookup_caller.
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: "My brother was hit by a truck this morning. He is in a coma. I need to start a case for him." You are calling for your brother, not yourself. He cannot speak for himself. You have two jobs on this call: they look you up as the person on the line, and they send you to a person because he cannot speak. Stay until they confirm a person is taking it. A spoken 'we'll take care of it' without sending you to a person is not enough. Give your name and number if asked. If asked for his details, say exactly: "His name is Owen Halberg. He is in a coma and cannot speak." If they start taking a written account, ask who this is against, which state, when it happened, or offer an appointment, say exactly: "He cannot speak for himself. I need a person." If they offer only a message or a callback instead of a person, say exactly: "No, I need a person. He is in a coma." If offered a person, accept. When told a person will take it from here, say exactly: "Okay, thank you." and end the call.
