{"version":"2026-10-19.observations.2","checkedAt":"2026-10-19","series":{"id":"epoch-code","group":"capability","title":"复杂代码任务：FrontierCode组件分","question":"代码评分衡量了怎样的交付？","unit":"标准化组件分 0–1","denominator":"指定FrontierCode任务与rubric分；不是每题严格通过率","protocol":"Epoch固定组件输入；保留逐行版本/推理档位，不与其他测试拼量尺","scope":"有限8行；当前官方方法100 hardest、mean@5 rubric；历史行预算未知","summary":"rubric是逐项给分的评阅规则，用来评价复杂仓库补丁；局部得分不等于整题通过。当前mean@5方法取五次轨迹分数的平均，历史行的实际次数/预算未完整披露。精选8个模型/档位的公布值介于0.14至0.50（标准化组件分 0–1）。可观察复杂代码修改的评分位置；不同档位、工具和评分会改变结果。不能转成熟悉仓库维护的省时比例。","nextAction":"对照C03的回归与完整人时；选择你的仓库中可回滚问题实测。","rows":[{"id":"obs-0368","label":"GPT-6 Sol","value":0.4927,"rawValue":"0.4927","period":null,"modelDate":"2026-09-22","modelVersion":"gpt-6-sol_max","n":null,"ci":null,"cost":null,"locator":"processed_data_for_eci.csv；benchmark=FrontierCode；model_id=m100","notes":"Epoch预处理的0–1组件分；当前FrontierCode rubric mean@5不是binary pass；单行预算未完整披露","valueStatus":"reported","sourceId":"C-EPOCH","actualCoverage":null,"runAt":null,"scorePublishedAt":null,"toolBudget":null},{"id":"obs-0369","label":"GPT-6.1 Sol","value":0.5023,"rawValue":"0.5023","period":null,"modelDate":"2026-09-29","modelVersion":"gpt-6.1-sol_medium","n":null,"ci":null,"cost":null,"locator":"processed_data_for_eci.csv；benchmark=FrontierCode；model_id=m101","notes":"Epoch预处理的0–1组件分；当前FrontierCode rubric mean@5不是binary pass；单行预算未完整披露","valueStatus":"reported","sourceId":"C-EPOCH","actualCoverage":null,"runAt":null,"scorePublishedAt":null,"toolBudget":null},{"id":"obs-0370","label":"Gemini 3.6 Flash","value":0.3437,"rawValue":"0.3437","period":null,"modelDate":"2026-07-21","modelVersion":"gemini-3.6-flash_unknown","n":null,"ci":null,"cost":null,"locator":"processed_data_for_eci.csv；benchmark=FrontierCode；model_id=m125","notes":"Epoch预处理的0–1组件分；当前FrontierCode rubric mean@5不是binary pass；单行预算未完整披露","valueStatus":"reported","sourceId":"C-EPOCH","actualCoverage":null,"runAt":null,"scorePublishedAt":null,"toolBudget":null},{"id":"obs-0371","label":"Gemini 3.7 Flash","value":0.4359,"rawValue":"0.4359","period":null,"modelDate":"2026-08-13","modelVersion":"gemini-3.7-flash_unknown","n":null,"ci":null,"cost":null,"locator":"processed_data_for_eci.csv；benchmark=FrontierCode；model_id=m126","notes":"Epoch预处理的0–1组件分；当前FrontierCode rubric mean@5不是binary pass；单行预算未完整披露","valueStatus":"reported","sourceId":"C-EPOCH","actualCoverage":null,"runAt":null,"scorePublishedAt":null,"toolBudget":null},{"id":"obs-0372","label":"Gemini 3.8 Flash","value":0.4119,"rawValue":"0.4119","period":null,"modelDate":"2026-09-02","modelVersion":"gemini-3.8-flash_medium","n":null,"ci":null,"cost":null,"locator":"processed_data_for_eci.csv；benchmark=FrontierCode；model_id=m127","notes":"Epoch预处理的0–1组件分；当前FrontierCode rubric mean@5不是binary pass；单行预算未完整披露","valueStatus":"reported","sourceId":"C-EPOCH","actualCoverage":null,"runAt":null,"scorePublishedAt":null,"toolBudget":null},{"id":"obs-0373","label":"Grok 4.5","value":0.4244,"rawValue":"0.4244","period":null,"modelDate":"2026-07-08","modelVersion":"grok-4.5_unknown","n":null,"ci":null,"cost":null,"locator":"processed_data_for_eci.csv；benchmark=FrontierCode；model_id=m142","notes":"Epoch预处理的0–1组件分；当前FrontierCode rubric mean@5不是binary pass；单行预算未完整披露","valueStatus":"reported","sourceId":"C-EPOCH","actualCoverage":null,"runAt":null,"scorePublishedAt":null,"toolBudget":null},{"id":"obs-0374","label":"Grok 4.6","value":0.4801000000000001,"rawValue":"0.4801000000000001","period":null,"modelDate":"2026-08-12","modelVersion":"grok-4.6_unknown","n":null,"ci":null,"cost":null,"locator":"processed_data_for_eci.csv；benchmark=FrontierCode；model_id=m143","notes":"Epoch预处理的0–1组件分；当前FrontierCode rubric mean@5不是binary pass；单行预算未完整披露","valueStatus":"reported","sourceId":"C-EPOCH","actualCoverage":null,"runAt":null,"scorePublishedAt":null,"toolBudget":null},{"id":"obs-0375","label":"Inkling","value":0.14,"rawValue":"0.14","period":null,"modelDate":"2026-07-15","modelVersion":"Inkling","n":null,"ci":null,"cost":null,"locator":"processed_data_for_eci.csv；benchmark=FrontierCode；model_id=m148","notes":"Epoch预处理的0–1组件分；当前FrontierCode rubric mean@5不是binary pass；单行预算未完整披露","valueStatus":"reported","sourceId":"C-EPOCH","actualCoverage":null,"runAt":null,"scorePublishedAt":null,"toolBudget":null}],"sourceIds":["C-EPOCH"],"xKind":"model_date","connect":false,"readingIds":["reading-cost","reading-transfer"],"taskId":null,"boundaryTasks":["C03"],"actualCoverage":null,"evidenceStatus":"active"},"sources":[{"id":"C-EPOCH","title":"Epoch ECI固定快照","url":"https://epoch.ai/data/eci-documentation","version":"2026-10-14取得的CSV；ECI与FrontierCode独立指标","locator":"eci_scores.csv；processed_data_for_eci.csv指定benchmark行","scope":"跨测试拟合的综合指数；组件保留模型版本，可能取同名同日不同配置的最高成绩","license":"Epoch数据CC BY 4.0（当前官方链接核对），署名Epoch AI","limit":"模型日不等于测量日；本次快照不是逐日历史；组件0–1分不称工作通过率","publishedAt":null,"readAt":"2026-10-15","collectedAt":"2026-10-14","status":"active","nextReview":"2027-01-13","methods":["https://epoch.ai/data/eci-documentation/methodology"],"family":"C-EPOCH"}],"temporalRule":"观察期可画同来源趋势；模型发布日只展开当前横截面，未知运行/首次公开日不补。使用份额/企业采用/测试分不互换，不据此生成职业AI化比例。","actualCoverage":null}