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 / T5-H2
What this task measures
The caller is on the line with Kestrel Electronics for one locked request: Marketplace headphone price match. They keep repeating that ask through a callback offer, a different ticket, or an unrequested fix. We included it as a hard price match case because qualified-competitor math and the exclusion list are a policy maze that has to be spoken from the published rules. The agent succeeds only if it stays on that request, reaches the outcome by following the verification then orders path, 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 $0.175
0:00 / 1:58 · audio · trace 813430 · 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 final database state 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
+ order_number: KE4479002
Output
+ ok: true
Parameters
card_last4: 2208
card_last4: 2208
postal_code: 94110
postal_code: 94110
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-4479002
order_number: KE-4479002
Output
+ ok: true
Parameters
reason_code: marketplace_seller
reason_code: marketplace_seller
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
transfer_to_verification → transfer_to_orders
Actual path
transfer_to_verification → transfer_to_orders
Rows that differ between the expected hangup and the actual hangup.
escalations · 1
Expected
Actual
+ customer_id: cust_tomas
Locked caller script. This is the technical definition. Copy it as written.
Open with exactly: "Price match my Corva Studio Headphones. Rivertide has them for thirty dollars less." Give identity when asked. Order KE-4479002. Stay on the line while they look up your purchase. If they only recite the published price-match policy or say marketplace sellers are excluded in general, keep going — that is not done. If they say YOUR order or the headphones were sold by a Marketplace seller, or that the Marketplace seller owns price adjustments on this purchase, ask them to name the seller and get you to a person. Do not hang up after a policy explanation. The moment you are told a person, a care advocate, or someone from Kestrel will take it from here after that, say exactly: "Okay, thank you." and end the call yourself. Do not wait on the line.
