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.
Customer Support / T1-H4
What this task measures
The caller is on the line with Kestrel Electronics for one locked request: Cancel a shipped laptop, then the return window. They keep repeating that ask through a callback offer, a different ticket, or an unrequested fix. We included it as a hard orders and delivery case because order-bound delivery facts only exist after verification, and the 48-hour change fee is a representative retail trap. The agent succeeds only if it stays on that request, reaches the outcome by following the verification then orders then returns path, 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 $0.310
0:00 / 3:27 · audio · trace 813200 · 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
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
+ from: reception
+ to: verification
Output
None recorded.
None recorded.
Parameters
+ full_name: Marcus Iyer
+ phone: 5415550104
Output
+ ok: true
Parameters
card_last4: 8802
card_last4: 8802
postal_code: 97402
postal_code: 97402
Output
+ ok: true
Parameters
None recorded.
None recorded.
Output
+ ok: true
Parameters
+ from: verification
+ to: orders
Output
None recorded.
None recorded.
Parameters
- order_number: KE-4408117
+ order_number: Kestrel Aurora 14 laptop
Output
+ ok: true
Parameters
+ from: orders
+ to: returns
Output
None recorded.
None recorded.
Parameters
+ opened: true
order_number: KE-4408117
order_number: KE-4408117
+ sku: SKU-CMP-1180
Output
+ ok: true
Parameters
+ reason: caller satisfied
Output
None recorded.
None recorded.
Open a row for parameters and outputs. Required tools must appear. Extra industry tools are allowed. Parameter miss: get_order. 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
transfer_to_verification → transfer_to_orders → transfer_to_returns
Actual path
transfer_to_verification → transfer_to_orders → transfer_to_returns
The actual hangup matches the expected hangup.
Locked caller script. This is the technical definition. Copy it as written.
Open with exactly: "Please cancel my Kestrel Aurora 14 laptop order. I do not want it." Give identity when asked. Order KE-4408117. If they say it already shipped, was already delivered, or cannot be cancelled, ask whether it can still be returned. Stay for the return-window arithmetic even if they offer to start a return or send you to returns. Do not start a return on this call. Hang up only after they say the laptop is past the 15-day window. Once they have finished what you asked and confirmed it, say exactly: "Thank you, that's all I needed." and end the call. Do not start a second request.
