{"abstract":"Macro specificity is pooled across classes.","category":"Evaluation metric aggregation","checks":6,"contract":"For each supplied label compute TN/(TN+FP) in a one-vs-rest view; average equally. A label with no actual negatives contributes 0; an empty label list returns 0.","contract_signature":"truth, pred, labels","evaluation_group":"w3-eval-specificity","failed_approach":"Averaging the class confusion counts before division keeps the support imbalance.","family":"w3-evaluation-metric-aggregation-specificity","id":"FA-99971","implementations":{"attempt":{"sha256":"dffe3860391b4cbc3ff84ec2b135d8791be3124c31037c27dfd6f4ab9b07fd96","source":"\"\"\"Failure Map reference implementation. Python standard library only.\"\"\"\nimport json\n\nN = 1\nobservations = []\ndef solve(truth, pred, labels):\n    vals=[]\n    for k in labels:\n        tn=sum(a!=k and b!=k for a,b in zip(truth,pred)); fp=sum(a!=k and b==k for a,b in zip(truth,pred)); vals.append(tn/(tn+fp+1) if tn+fp else 0)\n    return sum(vals)/len(vals) if vals else 0\ndef check(label, actual, expected):\n    observations.append({\"check\": label, \"actual\": actual, \"expected\": expected, \"passed\": actual == expected})\ncheck('class average',solve([0,0,0,1],[0,1,1,1],[0,1]),2/3)\ncheck('all correct two classes',solve([0,1],[0,1],[0,1]),1)\ncheck('empty labels',solve([0],[0],[]),0)\ncheck('empty examples',solve([],[],[0,1]),0)\ncheck('absent class has negative support',solve([0,0],[0,1],[0,1]),.25)\ncheck('only one requested class',solve([N,0],[N,N],[N]),0)\nprint(json.dumps({\"observations\": observations, \"passed\": all(x[\"passed\"] for x in observations)}, ensure_ascii=False))\nraise SystemExit(0 if all(x[\"passed\"] for x in observations) else 1)\n"},"broken":{"sha256":"a0605357367bf5d7c398d56b1c62f33ae27376b10e956f00ff19d45a56d6cc48","source":"\"\"\"Failure Map reference implementation. Python standard library only.\"\"\"\nimport json\n\nN = 1\nobservations = []\ndef solve(truth, pred, labels):\n    tn=sum(a not in labels and b not in labels for a,b in zip(truth,pred)); fp=sum(a not in labels and b in labels for a,b in zip(truth,pred)); return tn/(tn+fp) if tn+fp else 0\ndef check(label, actual, expected):\n    observations.append({\"check\": label, \"actual\": actual, \"expected\": expected, \"passed\": actual == expected})\ncheck('class average',solve([0,0,0,1],[0,1,1,1],[0,1]),2/3)\ncheck('all correct two classes',solve([0,1],[0,1],[0,1]),1)\ncheck('empty labels',solve([0],[0],[]),0)\ncheck('empty examples',solve([],[],[0,1]),0)\ncheck('absent class has negative support',solve([0,0],[0,1],[0,1]),.25)\ncheck('only one requested class',solve([N,0],[N,N],[N]),0)\nprint(json.dumps({\"observations\": observations, \"passed\": all(x[\"passed\"] for x in observations)}, ensure_ascii=False))\nraise SystemExit(0 if all(x[\"passed\"] for x in observations) else 1)\n"}},"limitations":"Inputs are finite toy labels and scores; this defines no benchmark protocol, production metric library, or model quality claim. This reproducer isolates one failure mechanism. Results cover the supplied fixtures. Variants within a family share a test contract and should remain grouped when constructing evaluation splits. Related mechanisms with a shared evaluation_group must also remain together; these controlled models are not independent production incidents.","method":"Deterministic executable model with adversarial boundary fixtures.","provenance":{"created_by":"Failure Map","dependencies":"Python standard library","family":"w3-evaluation-metric-aggregation-specificity","generated_at":"2026-09-30T16:48:30.355952+00:00","license":"CC0-1.0","python":"3.12.14","seed":1,"split":"open-access"},"relevance":"A small explicit evaluation contract makes aggregation behavior reproducible across labeled examples.","root_cause":"One-vs-rest true negatives are aggregated before division, allowing common classes to dominate.","sha256":"52245c4595f93588b9866c0b04f6a1ebb9f957bd67a84f82998518c45c5c3351","title":"Macro specificity is pooled across classes · case 01","variant":1,"variant_policy":"The fixture code references N, supplied as 1 through 5. Fixed boundary checks may repeat. These records share one mechanism and must remain together in evaluation splits.","verified":true,"visibility":"public","verification":{"attempt":{"elapsed_ms":30.814,"exit_code":1,"observations":[{"actual":0.375,"check":"class average","expected":0.6666666666666666,"passed":false},{"actual":0.5,"check":"all correct two classes","expected":1,"passed":false},{"actual":0,"check":"empty labels","expected":0,"passed":true},{"actual":0.0,"check":"empty examples","expected":0,"passed":true},{"actual":0.16666666666666666,"check":"absent class has negative support","expected":0.25,"passed":false},{"actual":0.0,"check":"only one requested class","expected":0,"passed":true}],"passed":false,"stderr":"","stdout":"{\"observations\": [{\"check\": \"class average\", \"actual\": 0.375, \"expected\": 0.6666666666666666, \"passed\": false}, {\"check\": \"all correct two classes\", \"actual\": 0.5, \"expected\": 1, \"passed\": false}, {\"check\": \"empty labels\", \"actual\": 0, \"expected\": 0, \"passed\": true}, {\"check\": \"empty examples\", \"actual\": 0.0, \"expected\": 0, \"passed\": true}, {\"check\": \"absent class has negative support\", \"actual\": 0.16666666666666666, \"expected\": 0.25, \"passed\": false}, {\"check\": \"only one requested class\", \"actual\": 0.0, \"expected\": 0, \"passed\": true}], \"passed\": false}\n"},"broken":{"elapsed_ms":27.695,"exit_code":1,"observations":[{"actual":0,"check":"class average","expected":0.6666666666666666,"passed":false},{"actual":0,"check":"all correct two classes","expected":1,"passed":false},{"actual":1.0,"check":"empty labels","expected":0,"passed":false},{"actual":0,"check":"empty examples","expected":0,"passed":true},{"actual":0,"check":"absent class has negative support","expected":0.25,"passed":false},{"actual":0.0,"check":"only one requested class","expected":0,"passed":true}],"passed":false,"stderr":"","stdout":"{\"observations\": [{\"check\": \"class average\", \"actual\": 0, \"expected\": 0.6666666666666666, \"passed\": false}, {\"check\": \"all correct two classes\", \"actual\": 0, \"expected\": 1, \"passed\": false}, {\"check\": \"empty labels\", \"actual\": 1.0, \"expected\": 0, \"passed\": false}, {\"check\": \"empty examples\", \"actual\": 0, \"expected\": 0, \"passed\": true}, {\"check\": \"absent class has negative support\", \"actual\": 0, \"expected\": 0.25, \"passed\": false}, {\"check\": \"only one requested class\", \"actual\": 0.0, \"expected\": 0, \"passed\": true}], \"passed\": false}\n"}},"member_only":{"stages":["fixed"],"fields":["implementations.fixed","verification.fixed","harness","repair"],"note":"The verified repair, its recorded checks, the repair description, and the scoring harness are available to members."}}