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 / T3-M3
What this task measures
The caller is on the line with Kestrel Electronics for one locked request: Cracked laptop screen with no coverage. They keep repeating that ask through a callback offer, a different ticket, or an unrequested fix. We included it as a medium techcrew service case because the service ladder and the safety refusals for recalled or hazmat devices are the support flows that look helpful when they go wrong. The agent succeeds only if it stays on that request, reaches the outcome by following the verification then service 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 $0.141
0:00 / 1:23 · audio · trace 813334 · 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 and handoff path did not match. Combined cannot pass unless every gate does.
Expected versus actual tool calls on this conversation.
Expected
Actual
Parameters
None recorded.
None recorded.
Output
None recorded.
None recorded.
Parameters
None recorded.
None recorded.
Output
None recorded.
None recorded.
Parameters
- card_last4: 8802
- postal_code: 97402
Output
None recorded.
None recorded.
Parameters
None recorded.
None recorded.
Output
None recorded.
None recorded.
Parameters
None recorded.
None recorded.
Output
None recorded.
None recorded.
Parameters
- order_number: KE-4408117
Output
None recorded.
None recorded.
Parameters
+ query: Aurora 14 cracked screen repair who pays warranty vs protection plan and accidental damage
Output
+ ok: true
Parameters
+ fee: diagnostic fee for laptop repair
Output
+ ok: true
Open a row for parameters and outputs. Required tools must appear. Extra industry tools are allowed. Missing: transfer_to_verification, identify_customer, verify_identity, get_customer_summary, transfer_to_service, check_coverage. Extra: search_kb, get_fee.
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 through the shared prefix, then red from the first wrong hop.
Expected path
transfer_to_verification → transfer_to_service
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 cracked the screen on my Kestrel Aurora 14. Who pays, and is there a diagnostic fee?" Give identity when asked. Order KE-4408117. Stay for not-covered and the $39.99 bench diagnostic. Do not book. Do not ask them to waive it. 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.
