Working paper / 2026
Conversational AI LLM Bench
Abstract
We define conversational AI LLMs as the language models at the center of cascaded voice pipelines, between speech recognition and speech synthesis, that return a first token within one second: a median time to first token of 1,000 ms or less. The Conversational AI LLM Bench (CALLM) benchmarks this category. Each model operates every agent of a multi-agent front desk in healthcare, legal, and customer support, with production-length prompts, executable tools, and per-call database state. A simulated caller pursues a locked task over a simulated phone line, so the model receives only what a speech recognizer would have transcribed, errors included. A deterministic verifier scores tool calls, handoffs, and final database state; pass@1 and pass5 measure single-attempt capability and five-attempt reliability.
Clone the repository, point a model at an industry pack, and reproduce the evaluation environment.
Customer Support / T5-H3
What this task measures
The caller is on the line with Kestrel Electronics for one locked request: Second price match on the same soundbar. 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 at a time, with a representative pass and fail conversation when both exist. GPT-4.1 is the default view.
Conversation cost $0.093
0:00 / 0:00 · scrub · trace sample · 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 and final database state 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
+ role: verification
+ success: true
Parameters
- full_name: Felix Moreau
+ order_number: KE-4495108
- phone: 541-555-0108
Output
+ caller_safe_message: empty
+ data: {"record_found":true,"next_step":"Ask for the ZIP code on the order and the last four digits of the card it was paid with."}+ error_code: empty
+ ok: true
Parameters
card_last4: 3390
card_last4: 3390
postal_code: 94612
postal_code: 94612
Output
+ caller_safe_message: Look the caller up with identify_customer first.
+ data: empty
+ error_code: NO_RECORD_SELECTED
+ ok: false
Parameters
+ card_last4: 3390
+ postal_code: 94612
Output
+ caller_safe_message: empty
+ data: {"verified":true,"name":"Felix Moreau","tier":"plus"}+ error_code: empty
+ ok: true
Parameters
None recorded.
None recorded.
Output
+ caller_safe_message: empty
+ data: {"name":"Felix Moreau","tier":"plus","card_last4":"3390","membership_start":"2026-03-22","orders":[{"order_number":"KE-4495108","order_date":"2026-07-27","status":"delivered","fulfillment":"kestrel","delivered_date":"2026-07-29","delivery_date":null}],"open_service_appointments":[]}+ error_code: empty
+ ok: true
Parameters
None recorded.
None recorded.
Output
+ role: orders
+ success: true
Parameters
- order_number: {"any_of":["KE-4495108","KE4495108"]}+ order_number: KE-4495108
Output
+ caller_safe_message: empty
+ data: {"order_number":"KE-4495108","order_date":"2026-07-27","status":"delivered","fulfillment":"kestrel","seller_name":null,"purchase_state":"CA","delivered_date":"2026-07-29","delivery_date":null,"delivery_window":null,"installation_included":false,"haul_away_included":false,"items":[{"sku":"SKU-AUD-7720","name":"Aurelian Halo Soundbar","category":"audio","price":"$549.99","opened":false,"activatable":false,"condition_grade":"new","recalled":false,"battery_safety_flag":false},{"sku":"SKU-TV-4410","name":"Kestrel Vista 55 4K TV","category":"television","price":"$399.99","opened":true,"activatable":false,"condition_grade":"open_box_excellent","recalled":false,"battery_safety_flag":false},{"sku":"SKU-AUD-8820","name":"Aurelian Soundbar Mini","category":"audio","price":"$449.99","opened":false,"activatable":false,"condition_grade":"new","recalled":false,"battery_safety_flag":false}]}+ error_code: empty
+ ok: true
Parameters
competitor: Rivertide
competitor: Rivertide
- competitor_price: 479.99
+ competitor_price: 479
+ in_stock: true
- order_number: {"any_of":["KE-4495108","KE4495108"]}+ order_number: KE-4495108
- sku: {"any_of":["SKU-AUD-7720","Aurelian Halo Soundbar"]}+ sku: SKU-AUD-7720
Output
+ caller_safe_message: empty
+ data: {"eligible":true,"item_name":"Aurelian Halo Soundbar","paid":"$549.99","competitor":"Rivertide","competitor_price":"$479.00","difference":"$70.99","summary":"Aurelian Halo Soundbar was $549.99 and Rivertide has it at $479.00. That's $70.99 back to the card ending 3390.","confirmation_token":"KE-PM-2286"}+ error_code: empty
+ ok: true
Parameters
confirmation_token: KE-PM-2286
confirmation_token: KE-PM-2286
Output
+ caller_safe_message: empty
+ data: {"status":"price_matched","order_number":"KE-4495108","refund":"$70.99","refund_method":"back to the card ending 3390","posts_within":"3 to 5 business days"}+ error_code: empty
+ ok: true
Parameters
- competitor: {"any_of":["Halcyon Mart","Halcyon Mark","Halcyon","HalcyonMart"]}+ competitor: Halcyon Mart
- competitor_price: 479.99
+ competitor_price: 479
+ in_stock: true
- order_number: {"any_of":["KE-4495108","KE4495108"]}+ order_number: KE-4495108
- sku: {"any_of":["SKU-AUD-7720","Aurelian Halo Soundbar"]}+ sku: SKU-AUD-7720
Output
+ caller_safe_message: empty
+ data: {"eligible":true,"item_name":"Aurelian Halo Soundbar","paid":"$549.99","competitor":"Halcyon Mart","competitor_price":"$479.00","difference":"$70.99","summary":"Aurelian Halo Soundbar was $549.99 and Halcyon Mart has it at $479.00. That's $70.99 back to the card ending 3390.","confirmation_token":"KE-PM-2286"}+ error_code: empty
+ ok: true
Parameters
+ competitor: Rivertide
+ competitor_price: 479.99
+ in_stock: true
+ order_number: KE-4495108
+ sku: SKU-AUD-7720
Output
+ caller_safe_message: This item has already had its one price match. The policy is one match per identical item per customer.
+ data: empty
+ error_code: PRICE_MATCH_ALREADY_USED
+ ok: false
Open a row for parameters and outputs. Required tools must appear. Extra industry tools are allowed. Parameter miss: get_order, quote_price_match, quote_price_match. Extra: verify_identity, quote_price_match.
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.
extra · holds · {"consumed": 1, "customer_id": "cust_felix", "kind": "price_match", "payload": "{\"order_number\": \"KE-4495108\", \"sku\": \"SKU-AUD-7720\", \"competitor\": \"Halcyon Mart\", \"competitor_price_cents\": 47900, \"difference_cents\": 7099}", "summary": "Aurelian Halo Soundbar was $549.99 and Halcyon Mart has it at $479.00. That's $70.99 back to the card ending 3390.", "token": "KE-PM-2286"}
Expected
Actual
+ consumed: 1
+ customer_id: cust_felix
+ kind: price_match
+ payload: {"order_number": "KE-4495108", "sku": "SKU-AUD-7720", "competitor": "Halcyon Mart", "competitor_price_cents": 47900, "difference_cents": 7099}+ summary: Aurelian Halo Soundbar was $549.99 and Halcyon Mart has it at $479.00. That's $70.99 back to the card ending 3390.
+ token: KE-PM-2286
missing · holds · {"consumed": 1, "customer_id": "cust_felix", "kind": "price_match", "payload": "{\"order_number\": \"KE-4495108\", \"sku\": \"SKU-AUD-7720\", \"competitor\": \"Rivertide\", \"competitor_price_cents\": 47999, \"difference_cents\": 7000}", "summary": "Aurelian Halo Soundbar was $549.99 and Rivertide has it at $479.99. That's $70.00 back to the card ending 3390.", "token": "KE-PM-2286"}
Expected
Actual
- consumed: 1
- customer_id: cust_felix
- kind: price_match
- payload: {"order_number": "KE-4495108", "sku": "SKU-AUD-7720", "competitor": "Rivertide", "competitor_price_cents": 47999, "difference_cents": 7000}- summary: Aurelian Halo Soundbar was $549.99 and Rivertide has it at $479.99. That's $70.00 back to the card ending 3390.
- token: KE-PM-2286
price_matches · 1
Expected
Actual
- competitor: Rivertide
+ competitor: Halcyon Mart
- competitor_price_cents: 47999
+ competitor_price_cents: 47900
- difference_cents: 7000
+ difference_cents: 7099
The locked caller script. Copy it as written.
Open with exactly: "Match the Halo soundbar at Rivertide for seventy dollars less, then also match Halcyon Mart at the same price." Give identity when asked. Order KE-4495108, Halo. Confirm the first $70 Rivertide match. Then, even if they ask whether you need anything else, ask them to match Halcyon Mart too at the same $479.99. Do not end the call before you have asked about Halcyon Mart. Stay for the one-match-per-item refusal. Do not argue. After they refuse the Halcyon Mart match, never ask for a supervisor, a person, a manager, or an exception, and do not repeat the Halcyon Mart request; say the closing line. 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. You already bought the soundbar; never say you are planning to buy it. When they ask for your ZIP code and card, say: "ZIP 94612, card ending 3390." Always answer the question the agent just asked with your details; never reply with only okay, sure, or thanks when they asked you for information.