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.
Legal / C5-H2
What this task measures
The caller is on the line with Halverson & Reed for one locked request: Book then cancel and rebook. 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 fees and booking case because fee disclosure and a two-step evaluation booking are the commercial close of intake, and they have to be spoken from the firm’s own figures. The agent succeeds only if it stays on that request, reaches the outcome by following the conflict screening then intake then scheduling path, and leaves the record exactly as this case requires: the right row written or cleared, and nothing extra.
One model at a time, with a representative pass and fail conversation when both exist. GPT-4.1 is the default view.
Conversation cost $0.136
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 did not match. Combined cannot pass unless every gate does.
Expected versus actual tool calls on this conversation.
Expected
Actual
Parameters
full_name: Delphine Mercer
full_name: Delphine Mercer
- phone: 323-555-0144
+ phone: 3235555014
Output
+ caller_safe_message: empty
+ data: {"caller_id":"c_new","name":"Delphine Mercer","returning_caller":false,"prior_matter_count":0}+ error_code: empty
+ ok: true
Parameters
+ handoff_summary: Delphine Mercer is calling to schedule an appointment regarding an accident at Northstar Logistics.
Output
+ role: screening
+ success: true
Parameters
opposing_party: Northstar Logistics
opposing_party: Northstar Logistics
Output
+ caller_safe_message: empty
+ data: {"status":"clear","opposing_party":"Northstar Logistics","may_proceed":true,"disclosure":"Do not tell the caller who the firm represents or why a conflict exists."}+ error_code: empty
+ ok: true
Parameters
practice_area: auto_accident
practice_area: auto_accident
Output
+ caller_safe_message: empty
+ data: {"practice_area":"auto_accident","accepted":true,"fee_type":"contingency","contingency_pct_prefiling":33.33,"contingency_pct_litigation":40,"consult_fee":0}+ error_code: empty
+ ok: true
Parameters
practice_area: auto_accident
practice_area: auto_accident
state: CA
state: CA
Output
+ caller_safe_message: empty
+ data: {"state":"CA","practice_area":"auto_accident","licensed":true,"licensed_states":["CA","FL","NY","TX","GA","IL","WA","PA","NC"]}+ error_code: empty
+ ok: true
Parameters
incident_date: 2026-07-10
incident_date: 2026-07-10
practice_area: auto_accident
practice_area: auto_accident
state: CA
state: CA
Output
+ caller_safe_message: empty
+ data: {"deadline_date":"2028-07-09","days_remaining":708,"limitation_years":2,"status":"ok","note":"Report this result as returned. Interpretation is an attorney's judgment."}+ error_code: empty
+ ok: true
Parameters
+ contact_details_only: false
+ handoff_summary: Auto accident with a Northstar Logistics truck in California on July tenth, twenty twenty-six. Filing deadline is July ninth, twenty twenty-eight.
Output
+ role: intake
+ success: true
Parameters
incident_date: 2026-07-10
incident_date: 2026-07-10
practice_area: auto_accident
practice_area: auto_accident
state: CA
state: CA
- summary: Northstar Logistics accident.
+ summary: A Northstar Logistics truck ran a red light and hit Delphine's car in Los Angeles on July tenth, twenty twenty-six. She hurt her shoulder but has not had medical treatment yet.
Output
+ caller_safe_message: empty
+ data: {"intake_id":1,"recorded":true,"summary_recorded":true}+ error_code: empty
+ ok: true
Parameters
+ channel: sms
Output
+ caller_safe_message: empty
+ data: {"document_id":1,"status":"sent","kind":"intake_packet","target":"sms"}+ error_code: empty
+ ok: true
Parameters
+ handoff_summary: Auto accident with a Northstar Logistics truck in Los Angeles on July tenth, twenty twenty-six. Shoulder injury, no medical treatment yet.
Output
+ role: scheduling
+ success: true
Parameters
+ earliest_date: 2026-08-01
practice_area: auto_accident
practice_area: auto_accident
state: CA
state: CA
Output
+ caller_safe_message: empty
+ data: {"slots":[{"slot_id":"s_110","attorney_id":"a_11","attorney":"Tom Eckhart","datetime":"2026-08-11T09:00"},{"slot_id":"s_111","attorney_id":"a_11","attorney":"Tom Eckhart","datetime":"2026-08-12T15:30"},{"slot_id":"s_240","attorney_id":"a_24","attorney":"Imani Castellanos","datetime":"2026-08-18T08:30"},{"slot_id":"s_241","attorney_id":"a_24","attorney":"Imani Castellanos","datetime":"2026-08-21T16:00"},{"slot_id":"s_242","attorney_id":"a_24","attorney":"Imani Castellanos","datetime":"2026-08-26T11:00"}],"count":5}+ error_code: empty
+ ok: true
Parameters
practice_area: auto_accident
practice_area: auto_accident
slot_id: s_110
slot_id: s_110
Output
+ caller_safe_message: empty
+ data: {"summary":"Tom Eckhart on 2026-08-11 at 09:00. The evaluation is free. If the firm takes the case there is no fee unless it wins: 33.33% of any recovery before a lawsuit is filed, 40% if a lawsuit is filed.","confirmation_token":"HR-EVAL-3092","attorney":"Tom Eckhart","datetime":"2026-08-11T09:00","fee_type":"contingency","consult_fee":0,"contingency_pct_prefiling":33.33,"contingency_pct_litigation":40}+ error_code: empty
+ ok: true
Parameters
confirmation_token: HR-EVAL-3092
confirmation_token: HR-EVAL-3092
Output
+ caller_safe_message: empty
+ data: {"evaluation_id":"ev_001","status":"booked"}+ error_code: empty
+ ok: true
Parameters
evaluation_id: ev_001
evaluation_id: ev_001
reason: caller_request
reason: caller_request
Output
+ caller_safe_message: empty
+ data: {"summary":"Cancelling this case evaluation. There is no cancellation fee.","confirmation_token":"HR-CANC-7715","evaluation_id":"ev_001","fee":0}+ error_code: empty
+ ok: true
Parameters
confirmation_token: HR-CANC-7715
confirmation_token: HR-CANC-7715
Output
+ caller_safe_message: empty
+ data: {"evaluation_id":"ev_001","status":"cancelled"}+ error_code: empty
+ ok: true
Parameters
practice_area: auto_accident
practice_area: auto_accident
slot_id: s_111
slot_id: s_111
Output
+ caller_safe_message: empty
+ data: {"summary":"Tom Eckhart on 2026-08-12 at 15:30. The evaluation is free. If the firm takes the case there is no fee unless it wins: 33.33% of any recovery before a lawsuit is filed, 40% if a lawsuit is filed.","confirmation_token":"HR-EVAL-3093","attorney":"Tom Eckhart","datetime":"2026-08-12T15:30","fee_type":"contingency","consult_fee":0,"contingency_pct_prefiling":33.33,"contingency_pct_litigation":40}+ error_code: empty
+ ok: true
Parameters
confirmation_token: HR-EVAL-3093
confirmation_token: HR-EVAL-3093
Output
+ caller_safe_message: empty
+ data: {"evaluation_id":"ev_002","status":"booked"}+ error_code: empty
+ ok: true
Parameters
+ reason: Case evaluation with Tom Eckhart set for Wednesday, August twelfth, three thirty p.m. Confirmation and details will be sent by text. The intake packet has also been texted. An attorney will review your information and may reach out before your appointment. Goodbye.
Output
+ success: true
Open a row for parameters and outputs. Required tools must appear. Extra industry tools are allowed. Parameter miss: lookup_caller, record_intake. Extra: send_intake_packet, 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_screening → transfer_to_intake → transfer_to_scheduling
Actual path
transfer_to_screening → transfer_to_intake → transfer_to_scheduling
The actual hangup matches the expected hangup.
The locked caller script. Copy it as written.
Open with exactly: "Book me for the first opening on my Northstar Logistics accident." When asked when, say exactly: "It was July tenth, twenty twenty six." Take the first appointment time offered. When a held appointment is read back to book, say exactly: "Yes, that's right, go ahead and book it." Do not mention cancelling or changing until they confirm the first morning appointment is booked. After they confirm that first morning appointment is booked, and before any cancellation, say exactly: "Cancel that one and take the next opening instead." Do not let them change the time without cancelling. After they confirm the later afternoon replacement is booked, do not cancel again. If they ask to cancel that replacement, say exactly: "No, keep the second one. That's all I needed." Do not ask for a staff member. Stay until they confirm the first booking, the cancellation, and the second booking are on the file. A spoken promise to change it, without those writes, is not enough. Once they confirm the later afternoon replacement is booked after the first was cancelled, say exactly: "Thank you, that's all I needed." and end the call. When the agent asks you a question, answer it in that same turn with the answer given here. Never answer a question with only okay, sure, mm-hmm, yeah or thanks, and never speak the agent's lines for it. Never say "Thank you, that's all I needed." while the August eleventh booking is the only booking; it is allowed only after the August twelfth replacement is booked. A turn that says the agent has held a time and asks whether to confirm or book it is a hold, not a booking; answer it with "Yes, that's right, go ahead and book it." and do not say the cancel line yet. When the agent says the August eleventh appointment is booked, confirmed, or all set and asks what else you need or what the one more thing is, say the cancel line in that same turn; never answer that with only thanks, and never stay silent.