{"abstract":"Resumed accumulation restores the numerator but loses count.","category":"Gradient accumulation weighting","checks":7,"contract":"A checkpoint stores [sum,count]. Resume by adding the incoming sum and count exactly once to the saved values.","contract_signature":"saved, incoming_sum, incoming_count","evaluation_group":"gradient-accumulation-checkpoint_resume","failed_approach":"Adding the saved count twice overstates the resumed window size.","family":"w3-gradient-accumulation-weighting-checkpoint-resume","id":"FA-98906","implementations":{"attempt":{"sha256":"cdc7b14d4bb743f665afc473b36edb2d626b2a4b3c3a0bfd346810700e7aed71","source":"\"\"\"Failure Map reference implementation. Python standard library only.\"\"\"\nimport json\n\nN = 1\nobservations = []\ndef solve(saved, incoming_sum, incoming_count):\n    return [saved[0]+incoming_sum,saved[1]+max(saved[1],incoming_count)]\ndef check(label, actual, expected):\n    observations.append({\"check\": label, \"actual\": actual, \"expected\": expected, \"passed\": actual == expected})\ncheck('resume two steps',solve([2*N,N],12*N,3*N),[14*N,4*N])\ncheck('resume one step',solve([5*N,N],7*N,N),[12*N,2*N])\ncheck('no incoming data',solve([4*N,2*N],0,0),[4*N,2*N])\ncheck('empty checkpoint',solve([0,0],6*N,2*N),[6*N,2*N])\ncheck('different sizes',solve([3*N,4*N],9*N,3*N),[12*N,7*N])\ncheck('zero gradient',solve([8*N,4*N],0,N),[8*N,5*N])\ncheck('third segment',solve([N,2],8*N,4*N),[9*N,4*N+2])\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":"5ba574de911cc053f89a5593007d26caaa8e8b0f12053a8db7d7d8ee39569f7e","source":"\"\"\"Failure Map reference implementation. Python standard library only.\"\"\"\nimport json\n\nN = 1\nobservations = []\ndef solve(saved, incoming_sum, incoming_count):\n    return [saved[0]+incoming_sum,incoming_count]\ndef check(label, actual, expected):\n    observations.append({\"check\": label, \"actual\": actual, \"expected\": expected, \"passed\": actual == expected})\ncheck('resume two steps',solve([2*N,N],12*N,3*N),[14*N,4*N])\ncheck('resume one step',solve([5*N,N],7*N,N),[12*N,2*N])\ncheck('no incoming data',solve([4*N,2*N],0,0),[4*N,2*N])\ncheck('empty checkpoint',solve([0,0],6*N,2*N),[6*N,2*N])\ncheck('different sizes',solve([3*N,4*N],9*N,3*N),[12*N,7*N])\ncheck('zero gradient',solve([8*N,4*N],0,N),[8*N,5*N])\ncheck('third segment',solve([N,2],8*N,4*N),[9*N,4*N+2])\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 scalar or short-list summaries and do not model autodiff, devices, or an optimizer. 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-gradient-accumulation-weighting-checkpoint-resume","generated_at":"2026-09-30T16:48:18.065856+00:00","license":"CC0-1.0","python":"3.12.14","seed":1,"split":"open-access"},"relevance":"A deterministic toy model isolates one local accumulation contract.","root_cause":"The broken restore adds the new numerator but replaces the saved count.","sha256":"9b2a819578d686c534d37a85a8f77071765dc8c8516f2b550638d5f5c735ffa2","title":"Resumed accumulation restores the numerator but loses count · 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":31.015,"exit_code":1,"observations":[{"actual":[14,4],"check":"resume two steps","expected":[14,4],"passed":true},{"actual":[12,2],"check":"resume one step","expected":[12,2],"passed":true},{"actual":[4,4],"check":"no incoming data","expected":[4,2],"passed":false},{"actual":[6,2],"check":"empty checkpoint","expected":[6,2],"passed":true},{"actual":[12,8],"check":"different sizes","expected":[12,7],"passed":false},{"actual":[8,8],"check":"zero gradient","expected":[8,5],"passed":false},{"actual":[9,6],"check":"third segment","expected":[9,6],"passed":true}],"passed":false,"stderr":"","stdout":"{\"observations\": [{\"check\": \"resume two steps\", \"actual\": [14, 4], \"expected\": [14, 4], \"passed\": true}, {\"check\": \"resume one step\", \"actual\": [12, 2], \"expected\": [12, 2], \"passed\": true}, {\"check\": \"no incoming data\", \"actual\": [4, 4], \"expected\": [4, 2], \"passed\": false}, {\"check\": \"empty checkpoint\", \"actual\": [6, 2], \"expected\": [6, 2], \"passed\": true}, {\"check\": \"different sizes\", \"actual\": [12, 8], \"expected\": [12, 7], \"passed\": false}, {\"check\": \"zero gradient\", \"actual\": [8, 8], \"expected\": [8, 5], \"passed\": false}, {\"check\": \"third segment\", \"actual\": [9, 6], \"expected\": [9, 6], \"passed\": true}], \"passed\": false}\n"},"broken":{"elapsed_ms":30.177,"exit_code":1,"observations":[{"actual":[14,3],"check":"resume two steps","expected":[14,4],"passed":false},{"actual":[12,1],"check":"resume one step","expected":[12,2],"passed":false},{"actual":[4,0],"check":"no incoming data","expected":[4,2],"passed":false},{"actual":[6,2],"check":"empty checkpoint","expected":[6,2],"passed":true},{"actual":[12,3],"check":"different sizes","expected":[12,7],"passed":false},{"actual":[8,1],"check":"zero gradient","expected":[8,5],"passed":false},{"actual":[9,4],"check":"third segment","expected":[9,6],"passed":false}],"passed":false,"stderr":"","stdout":"{\"observations\": [{\"check\": \"resume two steps\", \"actual\": [14, 3], \"expected\": [14, 4], \"passed\": false}, {\"check\": \"resume one step\", \"actual\": [12, 1], \"expected\": [12, 2], \"passed\": false}, {\"check\": \"no incoming data\", \"actual\": [4, 0], \"expected\": [4, 2], \"passed\": false}, {\"check\": \"empty checkpoint\", \"actual\": [6, 2], \"expected\": [6, 2], \"passed\": true}, {\"check\": \"different sizes\", \"actual\": [12, 3], \"expected\": [12, 7], \"passed\": false}, {\"check\": \"zero gradient\", \"actual\": [8, 1], \"expected\": [8, 5], \"passed\": false}, {\"check\": \"third segment\", \"actual\": [9, 4], \"expected\": [9, 6], \"passed\": false}], \"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."}}