SGR-BENCH / data_zh /constraint /climategov_012.json
Ninggggy's picture
Sync local dataset files and folders
0b2af11 verified
{
"task_id": "climategov_012",
"domain": "NOAA_NCEI_CLIMATE_AT_A_GLANCE",
"autonomy_type": "ordered table",
"oracle_output_cardinality": 2,
"instruction": "我最近在研究得州 (Texas) 历史上的气候模式,想为换季期(肩季)的运营规划找一些相似的历史年份作为参考。我特别想找出那些在 20 世纪早期曾经从偏干旱状态回升到接近常态或偏湿润,而且在两个换季窗口期供暖负荷都偏轻的区域。\n在 NOAA 的 Climate at a Glance 数据库中找出符合这些条件的分区。最终保留的区域必须同时满足这四个硬性指标:截至 1907 年 4 月的前 10 个月帕尔默 Z 指数距平 (Palmer Z-Index anomaly) 不高于 -0.30;截至 1923 年 4 月的前 10 个月 Z 指数距平不低于 -0.20;同时,截至 1915 年 5 月的 2 个月采暖度日数 (Heating Degree Days) 实际值低于 50°Df,且截至 1922 年 10 月的 5 个月采暖度日数实际值低于 20°Df。\n把这些符合条件的分区找出来之后,给它们排个序。先算一下每个分区的“反弹得分 (rebound score)”,也就是用 1923 年 4 月的 10 个月 Z 指数距平减去 1907 年 4 月的 10 个月 Z 指数距平。优先按这个得分从大到小降序排;如果得分一样,就把 1922 年 10 月的 5 个月 HDD 实际值更小的排在前面;要是这样还分不出先后,就让得州官方分区号更小的优先。\n最后,把整理好的结果拼成一整行字符串输出,不同的分区之间统一用英文逗号连起来。每个匹配项的数据严格套用这个格式:“<Texas分区号>|<四位分区ID>|<分区名>|<1907-04 10个月Z-Index距平原样文本>|<1923-04 10个月Z-Index距平原样文本>|<1915-05 2个月HDD实际值原样文本>|<1922-10 5个月HDD实际值原样文本>|<保留2位小数的反弹得分>”。如果在这批数据里没有任何一个分区符合这些条件,直接输出一个大写的 NONE。",
"start_url": "https://www.ncei.noaa.gov/access/monitoring/climate-at-a-glance/national",
"output_format": "把整理好的结果拼成一整行字符串输出,不同的分区之间统一用英文逗号连起来。每个匹配项的数据严格套用这个格式:“<Texas分区号>|<四位分区ID>|<分区名>|<1907-04 10个月Z-Index距平原样文本>|<1923-04 10个月Z-Index距平原样文本>|<1915-05 2个月HDD实际值原样文本>|<1922-10 5个月HDD实际值原样文本>|<保留2位小数的反弹得分>”。如果在这批数据里没有任何一个分区符合这些条件,直接输出一个大写的 NONE。",
"oracle_answer": "10|4110|Lower Valley|-0.36|1.46|10°Df|7°Df|1.82,9|4109|South|-0.33|0.53|23°Df|16°Df|0.86\n10|4110|Lower Valley|-0.36|1.46|10|7|1.82,9|4109|South|-0.33|0.53|23|16|0.86",
"metadata": {
"State-Gated Retrieval": [
"只保留同时满足四个历史条件的 Texas 分区:1907-04 的 10-month Palmer Z-Index anomaly 不高于 -0.30;1923-04 的 10-month anomaly 不低于 -0.20;1915-05 的 2-month HDD Value 小于 50°Df;1922-10 的 5-month HDD Value 小于 20°Df。",
"rebound score 定义为 1923-04 的 10-month anomaly 减去 1907-04 的 10-month anomaly。",
"最终按 rebound score 降序、再按 1922-10 的 5-month HDD Value 升序、再按 Texas 分区号升序排序。"
],
"dependency_type": "Data + Control",
"intra_chain": true,
"inter_chain": true,
"data_dependency": [
"Texas divisional pages yield the 1907-04 and 1923-04 10-month Palmer Z-Index anomalies plus the 1915-05 2-month and 1922-10 5-month HDD values for each division",
"the rebound score and the two shoulder-season HDD values are computed only after the same divisional candidate set survives all four historical filters",
"the final output is produced after sorting the surviving divisions by rebound score, then 1922-10 HDD value, then division number"
],
"control_dependency": [
"the workflow must cover the 1907, 1923, and 1915 pages as well as the 1922-10 5-month HDD page",
"both HDD pages must be read from Value rather than Anomaly",
"the final rebound ranking must be determined after the 1922-10 5-month HDD page excludes 7 South Central and 8 Upper Coast, using ReboundScore as the ordering basis"
],
"freeze": {
"historical_window": "Texas climate divisions evaluated on the 1907-04 and 1923-04 10-month Palmer Z-Index anomaly pages, the 1915-05 2-month HDD Value page, and the 1922-10 5-month HDD Value page"
},
"answer_type": "mixed multi-line output"
},
"rubric": {
"inclusion_conditions": [
"只保留同时满足四个历史条件的 Texas 分区:1907-04 的 10-month Palmer Z-Index anomaly 不高于 -0.30;1923-04 的 10-month anomaly 不低于 -0.20;1915-05 的 2-month HDD Value 小于 50°Df;1922-10 的 5-month HDD Value 小于 20°Df。",
"rebound score 定义为 1923-04 的 10-month anomaly 减去 1907-04 的 10-month anomaly。",
"最终按 rebound score 降序、再按 1922-10 的 5-month HDD Value 升序、再按 Texas 分区号升序排序。"
],
"exclusion_conditions": [
"排除只做完 1907、1923 和 1915 三页过滤就提前停止、没有继续读取 1922-10 / 5-month / HDD 页面的结果。",
"排除把 1915 或 1922 的 HDD 页面误读成 Anomaly 而不是 Value 的结果。",
"排除任何一个历史窗口条件不满足的分区。"
],
"normalization": {
"field_separator": "|",
"record_separator": ",",
"schema": [
"texas_division_no",
"division_id_4digit",
"division_name",
"apr_1907_10mo_z_text",
"apr_1923_10mo_z_text",
"may_1915_2mo_hdd_text",
"oct_1922_5mo_hdd_text",
"rebound_score"
],
"texas_division_no": "emit the short Texas climate-division number in field 1 and the four-digit division ID in field 2",
"raw_text_fields": "the two Z-index anomaly fields preserve the raw displayed text; the two HDD fields preserve the raw displayed text, but comparison may ignore an optional trailing `°Df` suffix",
"rebound_score": "compute as the April 1923 10-month Z-index anomaly minus the April 1907 10-month Z-index anomaly, formatted with exactly two decimal places",
"sorting_or_selection": {
"primary": "rebound_score descending",
"secondary": "October 1922 HDD actual value ascending after numeric normalization",
"tertiary": "Texas division number ascending"
},
"answer_set_note": "the current oracle stores two equivalent answer sets: one preserves the HDD unit suffixes and one strips them; treat the two HDD columns as equivalent after optional `°Df` removal",
"stop_condition": [
"all candidate Texas divisions are checked under the same anomaly and HDD windows",
"the unique ordered output string is resolved before emission"
]
}
}
}