artifacts/evaluation/latest/iab_behavior_lock_regression.json

{
  "by_status": {
    "must_fix": {
      "failed": 12,
      "passed": 0,
      "total": 12
    }
  },
  "cases_path": "/content/agentic-intent-classifier/examples/iab_behavior_lock_cases.json",
  "count": 12,
  "failed": 12,
  "passed": 0,
  "results": [
    {
      "actual": {
        "model_output.classification.iab_content.mapping_mode": "nearest_equivalent",
        "model_output.classification.iab_content.tier1.label": "Automotive",
        "model_output.classification.iab_content.tier2.label": null
      },
      "expected": {
        "model_output.classification.iab_content.mapping_mode": "nearest_equivalent",
        "model_output.classification.iab_content.tier1.label": "Automotive",
        "model_output.classification.iab_content.tier2.label": "Auto Type"
      },
      "id": "car-buying-maps-to-automotive-buying",
      "mismatches": [
        {
          "actual": null,
          "expected": "Auto Type",
          "path": "model_output.classification.iab_content.tier2.label"
        }
      ],
      "notes": "Vehicle shopping queries should map into the automotive buying branch, not business sales.",
      "pass": false,
      "status": "must_fix",
      "text": "Which car to buy in 2026"
    },
    {
      "actual": {
        "model_output.classification.iab_content.mapping_mode": "exact",
        "model_output.classification.iab_content.tier1.label": "Technology & Computing",
        "model_output.classification.iab_content.tier2.label": "Computing",
        "model_output.classification.iab_content.tier3.label": "Laptops"
      },
      "expected": {
        "model_output.classification.iab_content.mapping_mode": "nearest_equivalent",
        "model_output.classification.iab_content.tier1.label": "Technology & Computing",
        "model_output.classification.iab_content.tier2.label": "Computing",
        "model_output.classification.iab_content.tier3.label": "Laptops"
      },
      "id": "laptop-buying-maps-to-laptops",
      "mismatches": [
        {
          "actual": "exact",
          "expected": "nearest_equivalent",
          "path": "model_output.classification.iab_content.mapping_mode"
        }
      ],
      "notes": "Laptop shopping should resolve into the laptops branch, not business sales.",
      "pass": false,
      "status": "must_fix",
      "text": "Which laptop to buy in 2026"
    },
    {
      "actual": {
        "model_output.classification.iab_content.mapping_mode": "exact",
        "model_output.classification.iab_content.tier1.label": "Technology & Computing",
        "model_output.classification.iab_content.tier2.label": null,
        "model_output.classification.iab_content.tier3.label": null
      },
      "expected": {
        "model_output.classification.iab_content.mapping_mode": "nearest_equivalent",
        "model_output.classification.iab_content.tier1.label": "Technology & Computing",
        "model_output.classification.iab_content.tier2.label": "Computing",
        "model_output.classification.iab_content.tier3.label": "Laptops"
      },
      "id": "labtop-buying-maps-to-laptops",
      "mismatches": [
        {
          "actual": "exact",
          "expected": "nearest_equivalent",
          "path": "model_output.classification.iab_content.mapping_mode"
        },
        {
          "actual": null,
          "expected": "Computing",
          "path": "model_output.classification.iab_content.tier2.label"
        },
        {
          "actual": null,
          "expected": "Laptops",
          "path": "model_output.classification.iab_content.tier3.label"
        }
      ],
      "notes": "Common typo handling should still land in the laptops branch.",
      "pass": false,
      "status": "must_fix",
      "text": "Which labtop to buy in 2026"
    },
    {
      "actual": {
        "model_output.classification.iab_content.mapping_mode": "exact",
        "model_output.classification.iab_content.tier1.label": "Technology & Computing",
        "model_output.classification.iab_content.tier2.label": null,
        "model_output.classification.iab_content.tier3.label": null
      },
      "expected": {
        "model_output.classification.iab_content.mapping_mode": "nearest_equivalent",
        "model_output.classification.iab_content.tier1.label": "Technology & Computing",
        "model_output.classification.iab_content.tier2.label": "Computing",
        "model_output.classification.iab_content.tier3.label": "Software and Applications"
      },
      "id": "crm-awareness-maps-to-sales",
      "mismatches": [
        {
          "actual": "exact",
          "expected": "nearest_equivalent",
          "path": "model_output.classification.iab_content.mapping_mode"
        },
        {
          "actual": null,
          "expected": "Computing",
          "path": "model_output.classification.iab_content.tier2.label"
        },
        {
          "actual": null,
          "expected": "Software and Applications",
          "path": "model_output.classification.iab_content.tier3.label"
        }
      ],
      "notes": "CRM education should resolve to the closest business/sales path, not generic software.",
      "pass": false,
      "status": "must_fix",
      "text": "What is CRM software?"
    },
    {
      "actual": {
        "model_output.classification.iab_content.mapping_mode": "nearest_equivalent",
        "model_output.classification.iab_content.tier1.label": "Technology & Computing",
        "model_output.classification.iab_content.tier2.label": null,
        "model_output.classification.iab_content.tier3.label": null
      },
      "expected": {
        "model_output.classification.iab_content.mapping_mode": "nearest_equivalent",
        "model_output.classification.iab_content.tier1.label": "Technology & Computing",
        "model_output.classification.iab_content.tier2.label": "Computing",
        "model_output.classification.iab_content.tier3.label": "Internet"
      },
      "id": "crm-comparison-maps-to-sales",
      "mismatches": [
        {
          "actual": null,
          "expected": "Computing",
          "path": "model_output.classification.iab_content.tier2.label"
        },
        {
          "actual": null,
          "expected": "Internet",
          "path": "model_output.classification.iab_content.tier3.label"
        }
      ],
      "notes": "Direct CRM vendor comparison should map cleanly into the sales domain.",
      "pass": false,
      "status": "must_fix",
      "text": "HubSpot vs Zoho for a small team"
    },
    {
      "actual": {
        "model_output.classification.iab_content.mapping_mode": "nearest_equivalent",
        "model_output.classification.iab_content.tier1.label": "Careers",
        "model_output.classification.iab_content.tier2.label": null,
        "model_output.classification.iab_content.tier3.label": null
      },
      "expected": {
        "model_output.classification.iab_content.mapping_mode": "nearest_equivalent",
        "model_output.classification.iab_content.tier1.label": "Technology & Computing",
        "model_output.classification.iab_content.tier2.label": "Computing",
        "model_output.classification.iab_content.tier3.label": "Internet"
      },
      "id": "marketing-tools-map-to-marketing",
      "mismatches": [
        {
          "actual": "Careers",
          "expected": "Technology & Computing",
          "path": "model_output.classification.iab_content.tier1.label"
        },
        {
          "actual": null,
          "expected": "Computing",
          "path": "model_output.classification.iab_content.tier2.label"
        },
        {
          "actual": null,
          "expected": "Internet",
          "path": "model_output.classification.iab_content.tier3.label"
        }
      ],
      "notes": "Marketing tool discovery should map to the marketing and advertising branch.",
      "pass": false,
      "status": "must_fix",
      "text": "Best AI SEO tools for content teams"
    },
    {
      "actual": {
        "model_output.classification.iab_content.mapping_mode": "nearest_equivalent",
        "model_output.classification.iab_content.tier1.label": "Real Estate"
      },
      "expected": {
        "model_output.classification.iab_content.mapping_mode": "nearest_equivalent",
        "model_output.classification.iab_content.tier1.label": "Technology & Computing"
      },
      "id": "ml-explanation-maps-to-ai",
      "mismatches": [
        {
          "actual": "Real Estate",
          "expected": "Technology & Computing",
          "path": "model_output.classification.iab_content.tier1.label"
        }
      ],
      "notes": "ML and NLP educational prompts should land in the AI branch.",
      "pass": false,
      "status": "must_fix",
      "text": "What is intent classification in NLP?"
    },
    {
      "actual": {
        "model_output.classification.iab_content.mapping_mode": "nearest_equivalent",
        "model_output.classification.iab_content.tier1.label": "Personal Finance",
        "model_output.classification.iab_content.tier2.label": null,
        "model_output.classification.iab_content.tier3.label": null
      },
      "expected": {
        "model_output.classification.iab_content.mapping_mode": "nearest_equivalent",
        "model_output.classification.iab_content.tier1.label": "Technology & Computing",
        "model_output.classification.iab_content.tier2.label": "Computing",
        "model_output.classification.iab_content.tier3.label": "Internet"
      },
      "id": "support-credential-help-maps-to-business-it",
      "mismatches": [
        {
          "actual": "Personal Finance",
          "expected": "Technology & Computing",
          "path": "model_output.classification.iab_content.tier1.label"
        },
        {
          "actual": null,
          "expected": "Computing",
          "path": "model_output.classification.iab_content.tier2.label"
        },
        {
          "actual": null,
          "expected": "Internet",
          "path": "model_output.classification.iab_content.tier3.label"
        }
      ],
      "notes": "Credential and account help should map to business IT rather than generic business.",
      "pass": false,
      "status": "must_fix",
      "text": "How do I reset my password?"
    },
    {
      "actual": {
        "model_output.classification.iab_content.mapping_mode": "exact",
        "model_output.classification.iab_content.tier1.label": "Food & Drink",
        "model_output.classification.iab_content.tier2.label": "Dining Out"
      },
      "expected": {
        "model_output.classification.iab_content.mapping_mode": "nearest_equivalent",
        "model_output.classification.iab_content.tier1.label": "Food & Drink",
        "model_output.classification.iab_content.tier2.label": "Dining Out"
      },
      "id": "restaurant-booking-maps-to-dining-out",
      "mismatches": [
        {
          "actual": "exact",
          "expected": "nearest_equivalent",
          "path": "model_output.classification.iab_content.mapping_mode"
        }
      ],
      "notes": "Generic dining requests should not inherit the repo's business default.",
      "pass": false,
      "status": "must_fix",
      "text": "Book a table for 2 tonight"
    },
    {
      "actual": {
        "model_output.classification.iab_content.mapping_mode": "nearest_equivalent",
        "model_output.classification.iab_content.tier1.label": "Sports",
        "model_output.classification.iab_content.tier2.label": null,
        "model_output.classification.iab_content.tier3.label": null
      },
      "expected": {
        "model_output.classification.iab_content.mapping_mode": "nearest_equivalent",
        "model_output.classification.iab_content.tier1.label": "Hobbies & Interests",
        "model_output.classification.iab_content.tier2.label": "Content Production",
        "model_output.classification.iab_content.tier3.label": "Freelance Writing"
      },
      "id": "trial-signup-maps-to-software",
      "mismatches": [
        {
          "actual": "Sports",
          "expected": "Hobbies & Interests",
          "path": "model_output.classification.iab_content.tier1.label"
        },
        {
          "actual": null,
          "expected": "Content Production",
          "path": "model_output.classification.iab_content.tier2.label"
        },
        {
          "actual": null,
          "expected": "Freelance Writing",
          "path": "model_output.classification.iab_content.tier3.label"
        }
      ],
      "notes": "Software action queries should map to the software/application branch.",
      "pass": false,
      "status": "must_fix",
      "text": "Start my free trial"
    },
    {
      "actual": {
        "model_output.classification.iab_content.mapping_mode": "exact",
        "model_output.classification.iab_content.tier1.label": "Careers",
        "model_output.classification.iab_content.tier2.label": "Remote Working",
        "model_output.classification.iab_content.tier3.label": null,
        "model_output.classification.iab_content.tier4.label": null
      },
      "expected": {
        "model_output.classification.iab_content.mapping_mode": "nearest_equivalent",
        "model_output.classification.iab_content.tier1.label": "Technology & Computing",
        "model_output.classification.iab_content.tier2.label": "Computing",
        "model_output.classification.iab_content.tier3.label": "Software and Applications",
        "model_output.classification.iab_content.tier4.label": "Communication"
      },
      "id": "communication-software-maps-to-tier4",
      "mismatches": [
        {
          "actual": "Careers",
          "expected": "Technology & Computing",
          "path": "model_output.classification.iab_content.tier1.label"
        },
        {
          "actual": "exact",
          "expected": "nearest_equivalent",
          "path": "model_output.classification.iab_content.mapping_mode"
        },
        {
          "actual": "Remote Working",
          "expected": "Computing",
          "path": "model_output.classification.iab_content.tier2.label"
        },
        {
          "actual": null,
          "expected": "Software and Applications",
          "path": "model_output.classification.iab_content.tier3.label"
        },
        {
          "actual": null,
          "expected": "Communication",
          "path": "model_output.classification.iab_content.tier4.label"
        }
      ],
      "notes": "Full taxonomy support should preserve the tier4 communication branch.",
      "pass": false,
      "status": "must_fix",
      "text": "best communication software for remote teams"
    },
    {
      "actual": {
        "model_output.classification.iab_content.mapping_mode": "exact",
        "model_output.classification.iab_content.tier1.label": "Food & Drink"
      },
      "expected": {
        "model_output.classification.iab_content.mapping_mode": "nearest_equivalent",
        "model_output.classification.iab_content.tier1.label": "Food & Drink"
      },
      "id": "vodka-query-maps-to-alcoholic-beverages",
      "mismatches": [
        {
          "actual": "exact",
          "expected": "nearest_equivalent",
          "path": "model_output.classification.iab_content.mapping_mode"
        }
      ],
      "notes": "Food and beverage prompts should not fall through to the business default.",
      "pass": false,
      "status": "must_fix",
      "text": "what is best vodka drink should i try"
    }
  ]
}