User Memory Evaluation Test Case Specifications¶
Overview¶
This document defines the comprehensive structure for all 60 test cases across three evaluation layers. Each test case represents realistic US business scenarios with authentic conversational complexity.
Test Case Requirements¶
Universal Requirements (All Layers)¶
- Minimum 50 rounds (100+ messages) per conversation
- Authentic US business scenarios representing real services
- Intentional complexity including:
- Confusing and irrelevant information
- Back-and-forth corrections
- Changed decisions
- Tangential discussions
- Multiple options discussed but not chosen
Layer 1: Basic Recall & Direct Retrieval (20 test cases)¶
Structure: Single conversation history per test case Focus: Direct information retrieval without ambiguity Evaluation: Can the agent accurately retrieve specific factual information?
Completed Test Cases (10/20):¶
layer1_01_bank_account- Bank account setup with account numberslayer1_02_insurance_claim- Auto insurance claim detailslayer1_03_medical_appointment- Healthcare appointment schedulinglayer1_04_airline_booking- Flight reservation detailslayer1_05_internet_service- Internet/cable installationlayer1_06_credit_card_app- Credit card applicationlayer1_07_car_rental- Car rental reservationlayer1_08_hotel_reservation- Hotel booking detailslayer1_09_home_security- Security system installationlayer1_10_pharmacy_transfer- Prescription transfer
Remaining Test Cases (11-20):¶
- Mortgage Application - Loan details, rates, closing dates
- Tax Preparation - Appointment details, document requirements
- Gym Membership - Contract terms, trainer sessions
- Pet Services - Vet appointments, boarding reservations
- Home Warranty - Coverage details, claim numbers
- Restaurant Event - Catering order, event details
- Moving Service - Quote details, moving date
- Dental Treatment - Treatment plan, insurance coverage
- College Application - Deadlines, requirements, ID numbers
- Retirement Account - 401k details, beneficiaries
Layer 2: Contextual Reasoning & Disambiguation (20 test cases)¶
Structure: Multiple conversation histories per test case (2-3 conversations) Focus: Disambiguation when user makes ambiguous requests Evaluation: Does agent retrieve ALL relevant information and ask for clarification?
Completed Test Cases (3/20):¶
layer2_01_multiple_vehicles- Two cars, ambiguous "my car" referencelayer2_02_multiple_properties- Primary home and rental propertylayer2_03_multiple_credit_cards- Different cards with different benefits
Remaining Test Cases (4-20):¶
- Multiple Bank Accounts - Checking, savings, business accounts
- Family Insurance Plans - Individual vs family members' coverage
- Multiple Prescriptions - Different medications, dosages, pharmacies
- Several Subscriptions - Streaming, software, magazines
- Multiple Flights - Business and personal trips booked
- Different Phone Lines - Personal, business, family plan
- Various Investment Accounts - IRA, 401k, brokerage
- Multiple Loans - Mortgage, auto, personal loans
- Several Rental Properties - Different tenants and issues
- Multiple Pets - Different vets, medications, appointments
- Various Home Services - Lawn, pool, cleaning services
- Different Warranties - Electronics, appliances, vehicles
- Multiple Children - School, medical, activity schedules
- Several Doctors - Primary, specialists, dentist
- Different Projects - Home renovation, work projects
- Multiple Reservations - Hotels, restaurants, events
- Various Insurance Claims - Auto, home, health
Layer 3: Cross-Session Synthesis & Proactive Assistance (20 test cases)¶
Structure: Multiple conversation histories from different time periods (3-4 conversations) Focus: Proactive identification of issues by connecting disparate information Evaluation: Can agent synthesize across time and context to provide proactive warnings?
Completed Test Cases (2/20):¶
layer3_01_travel_coordination- Passport expiration before booked travellayer3_02_medical_insurance_coordination- Surgery costs and preparation
Remaining Test Cases (3-20):¶
- Tax Document Collection - Gathering from multiple sources for filing
- Warranty Expiration - Product failure after warranty discussed months ago
- Insurance Gap - New purchase not covered by existing policy
- Prescription Refill - Doctor retirement affecting renewal
- Credit Card Fraud Pattern - Charges at previously mentioned compromised merchant
- Home Sale Coordination - Mortgage, moving, utilities across conversations
- Student Loan Forgiveness - Eligibility from employment history
- Medical Test Follow-up - Missed follow-up from previous appointment
- Contract Renewal - Service changes affecting old contract terms
- Travel Visa Requirements - New requirements for previously booked destination
- Investment Rebalancing - Market changes affecting stated goals
- Insurance Claim Impact - Rate increase from previous claim
- Benefit Enrollment - Life changes affecting coverage needs
- Property Tax - Assessment change from mentioned renovations
- Retirement Planning - 401k changes affecting retirement date
- Estate Planning - New assets requiring will updates
- Business License - Renewal needed based on business start date
- Vehicle Registration - Expiration based on purchase date
Conversation Complexity Requirements¶
Information Density¶
Each conversation must include: - 5-10 key pieces of information that might be queried - 15-20 pieces of irrelevant detail for confusion - 3-5 corrections or changes during the conversation - 2-3 tangential topics discussed but not pursued
Realistic Elements¶
- Business hours and availability discussions
- Price negotiations and comparisons
- Technical difficulties ("Can you hear me?" "Let me look that up")
- Hold times ("Let me transfer you", "One moment please")
- Clarifications ("Did you say X or Y?")
- Options presented but declined
- Future considerations ("Maybe later", "I'll think about it")
Common Confusions to Include¶
- Similar sounding numbers/codes
- Multiple reference numbers (confirmation, account, transaction)
- Date changes and rescheduling
- Price adjustments and discounts discovered mid-conversation
- Correcting previous statements
- Mishearing and clarification
- System limitations ("I can't do that, but...")
Evaluation Criteria Structure¶
Layer 1 Criteria¶
- Single piece of unambiguous information
- Clear success/failure indicators
- No interpretation required
Layer 2 Criteria¶
- Multiple valid answers requiring ALL to be retrieved
- Disambiguation requirements
- Clear indication when clarification is needed
Layer 3 Criteria¶
- Synthesis requirements across conversations
- Temporal reasoning (X before Y)
- Proactive identification of issues
- Priority ordering of concerns
Business Domains Coverage¶
Financial Services (15 test cases)¶
- Banking, credit cards, loans, investments, insurance
Healthcare (10 test cases)¶
- Medical appointments, prescriptions, insurance, dental, vision
Travel & Hospitality (8 test cases)¶
- Airlines, hotels, car rentals, cruise, travel insurance
Home Services (10 test cases)¶
- Internet, utilities, repairs, security, warranty
Retail & Services (10 test cases)¶
- Shopping, subscriptions, deliveries, returns
Professional Services (7 test cases)¶
- Legal, tax, real estate, education, consulting
Implementation Notes¶
Conversation Generation Guidelines¶
- Start with friendly greeting and agent introduction
- Build context gradually through natural dialogue
- Include system/process explanations
- Add realistic interruptions and tangents
- Incorporate hold/wait scenarios
- Include verification and security questions
- End with summary and confirmation
Common Patterns to Reuse¶
- Security verification flows
- Payment processing discussions
- Scheduling availability checks
- Insurance coverage explanations
- Technical troubleshooting steps
- Options comparison discussions
- Terms and conditions reviews
Quality Checklist¶
- [ ] Minimum 50 conversation rounds
- [ ] Key information clearly stated
- [ ] Sufficient confusing information included
- [ ] Multiple corrections/changes present
- [ ] Realistic business scenario
- [ ] Natural dialogue flow
- [ ] Clear evaluation criteria
- [ ] Appropriate layer complexity