Deliver INTERIM_100 R4bc scores and paired step contrasts with preserved failures

This commit is contained in:
Frank14f committed 2026-11-17 04:14:35 +08:00
1 parent a326870216
commit 1eb394799f
19 files changed
+568126 -4

No files matched your search

File diff suppressed because it is too large. Load diff
@@ -1,8 +1,8 @@
{
"stage": "predict",
"cell": 14,
"arm": "Q",
"post_cpu_settled": 312.80652499999997,
"stage": "score",
"cell": null,
"arm": null,
"post_cpu_settled": 317.177371,
"reserved_active_cpu": 1800,
"post_cap": 14400,
"interim_cap": 7200,
@@ -0,0 +1,18 @@
{
"status": "PASS",
"checked_unix": 1794860072.6684394,
"scope": "INTERIM_100",
"original_frozen_files_unchanged": 405,
"source_pins_unchanged": true,
"prediction_entries_hashed": 4968,
"new_prediction_guards_pass": 35,
"score_guard_pass": true,
"scored_rows": 4968,
"unique_parts": 552,
"unique_components": 204,
"undefined_metric_fields": 0,
"score_summary_sha256": "800df87374ec975f7e5e17d65a6754dfbd25771df886bc063f4f3853c73d6b0e",
"part_scores_sha256": "8577fbd8c48a905984e751e9880f05f5ee4754e6e13e41c9ec77bf60f3f80a1d",
"final_full_exact_match": "PENDING_FULL_RUN_NOT_CLAIMED",
"no_retraining_or_reprediction_in_audit": true
}
@@ -0,0 +1,14 @@
{
"observed_at": "2026-11-17T04:11:59+08:00",
"service": "meltmind-r4bc-interim100-score-20261117.service",
"active_state": "active",
"sub_state": "running",
"coordinator_pid": 536613,
"coordinator_cpu": 1,
"worker_pid": 536623,
"worker_cpu": 0,
"parts_scored_at_observation": 189,
"prediction_freeze_commit": "a326870216f34e1e38d90319228e92846838675b",
"prediction_freeze_sha256": "db2ceca68f0eedb94f031c0deb958df234f934fa4534ea444423c4668af3de1e",
"source_functions_unchanged": true
}
@@ -0,0 +1,8 @@
{
"status": "COMPLETE",
"scope": "INTERIM_100",
"post_cpu_seconds": 346.870387,
"remaining_post_cpu_seconds": 14053.129613,
"coordinator_cpu_seconds": 3.413776,
"next": "FIXED_RESULT_DELIVERY"
}
@@ -0,0 +1,5 @@
{
"pid": 536623,
"unix": 1794859909.1557975,
"guard_pid": 536613
}
Whitespace-only changes.
@@ -0,0 +1,20 @@
{
"status": "PASS",
"returncode": 0,
"stop_reason": null,
"cpu_seconds": 29.693016,
"wall_seconds": 29.991380218009,
"peak_tree_rss_bytes": 165539840,
"peak_process_count": 1,
"monitor_errors": [],
"max_poll_gap_seconds": 0.9280491379904561,
"reaped_orphan_cpu_seconds": 0.0,
"limits": {
"cpu_seconds": 1800,
"wall_seconds": 3600,
"rss_bytes": 8589934592,
"disk_bytes": 2147483648
},
"retry": false,
"gpu_visible": false
}
@@ -0,0 +1,16 @@
{
"command": [
"/home/frank14f/optane/MeltMind/artifacts/gre24-reproduction-2026-11-15/environment/venv/bin/python",
"tools/r4bc_interim.py",
"worker",
"--mode",
"score"
],
"limits": {
"cpu_seconds": 1800,
"wall_seconds": 3600,
"rss_bytes": 8589934592,
"disk_bytes": 2147483648
},
"unix": 1794859909.15151
}
Whitespace-only changes.
@@ -0,0 +1,5 @@
{
"pid": 536613,
"unix": 1794859908.495007,
"scope": "INTERIM_100"
}
File diff suppressed because it is too large. Load diff
@@ -0,0 +1,4 @@
{
"parts": 552,
"cpu_seconds": 27.046398415
}
File diff suppressed because it is too large. Load diff
File diff suppressed because it is too large. Load diff
@@ -0,0 +1,18 @@
{
"commit": "a326870216f34e1e38d90319228e92846838675b",
"objects": {
"a326870216f34e1e38d90319228e92846838675b": "dHJlZSBlODMzMzc2OWUwYzhlNmJlOGFlZmYxZTE3MDAwZTVmMDhhNDNiMDgwCnBhcmVudCAxZjVjZTkxM2JjZDQyNDI5MDE2MmNlYjhkOTRlNmZhNGNmZjk4YTEyCmF1dGhvciBGcmFuazE0ZiA8MTUxNTQ0NDMxNEBxcS5jb20+IDE3OTQ4NTk4NzMgKzA4MDAKY29tbWl0dGVyIEZyYW5rMTRmIDwxNTE1NDQ0MzE0QHFxLmNvbT4gMTc5NDg1OTg3MyArMDgwMAoKRnJlZXplIGFsbCBJTlRFUklNXzEwMCBwcmVkaWN0aW9ucyBiZWZvcmUgY2FjaGVkLXRhcmdldCBzY29yaW5nCg==",
"e8333769e0c8e6be8aeff1e17000e5f08a43b080": "MTAwNjQ0IEFHRU5UUy5tZACNn/8WP2Z5T3zdQ1XERgZgRRqGmTEwMDY0NCBSRUFETUUubWQAnV/cqOQlQuso9DckOIaiq4uPaJs0MDAwMCBhcnRpZmFjdHMAJMaLBqULL0JSZwOZGwQOmjih0Jk0MDAwMCBkb2NzAOSVV6pFwwF4WW0432U2svUspt8aNDAwMDAgcmVmZXJlbmNlcwCPIpjiGLKeB1d70up/JOCwPiCndTQwMDAwIHJldmlldwBknNsxy/leFsnjoy8KZ976cHeynzQwMDAwIHRlc3RzAELYeKnQNGDLL6ePXWQ2Z1t3Lm37NDAwMDAgdG9vbHMAmPBaoH0cU4xoaFf9ZAL7FM5FfZU=",
"24c68b06a50b2f42526703991b040e9a38a1d099": "NDAwMDAgY29tYmluZWQtc3RhdHVzLTIwMjYtMTEtMTYANxEM7/rOrTMDZbV0g7Iq+WBgFEY0MDAwMCBncmUyNC1yZXByb2R1Y3Rpb24tMjAyNi0xMS0xNQBUkxknxWfm4LorYlJ/EvAwmFHsUDQwMDAwIGdyZTI0LXY0LXJlcHJvZHVjdGlvbi0yMDI2LTExLTE2AOjIqk+6n9o8gSGqd9CxTMbTvnWCNDAwMDAgZ3JlMjQtdjQtc3RvcmVkLXNjb3JlLWRpc3BsYXktMjAyNi0xMS0xNgCy182jihWEaxke9YBv34D6j9N4EDQwMDAwIGdyZTI0LXY0LXRlcm1pbmFsLWF1ZGl0LTIwMjYtMTEtMTYAq44U1Xxa6vZ3vEN5lw68jqyK0g00MDAwMCBpbnRha2UtMjAyNi0xMS0xMwBN+r81FwB3vJJg1WwoKaLuLN5GHjQwMDAwIG1lbHRhc3Npc3QtdjQtdGVybWluYWwtYXVkaXQtMjAyNi0xMS0xNgCjSf+ZI4ERKdI0XE/8K6EdUyYBVDQwMDAwIG1vbGRmbG93LXRpbWUtbWFwLTIwMjYtMTEtMTUAUi5OJ+/+Ko7DxPygpm/q2GsctoQ0MDAwMCBuYWl2ZS1kb21haW4tY29udGludWF0aW9uLTIwMjYtMTEtMTUAYXXhvmNErRoSCo+EuwlYn1/lbVUxMDA2NDQgbmFpdmUtZG9tYWluLXB1YmxpY2F0aW9uLTIwMjYtMTEtMTUuanNvbgDS920VDIngo8iBpXYdsdA6m8wCXTQwMDAwIG5haXZlLWxvZy1qb2luLTIwMjYtMTEtMTUATNHZcBJtcsJMDQYsY8iCws9LRKQxMDA2NDQgbmFpdmUtbG9nLXJldmlldy0yMDI2LTExLTE1LnppcAA9RHuBsHngrrtXJf/GwUECliX1WjEwMDY0NCBuYWl2ZS1sb2ctcmV2aWV3LXNoYXJlZC0yMDI2LTExLTE1Lmpzb24AHMxaEA8hoHUM5l+qqIvW4lllRsI0MDAwMCBuYWl2ZS1vcmRlcmluZy0yMDI2LTExLTE1AAVOMC2WfLASbS8y7Z/ak/hPBJotNDAwMDAgcjRhLWFjY2VwdGFuY2UtMjAyNi0xMS0xNgDSSCbnt9+r6Inmm6a5IuNfLmNpkjQwMDAwIHI0YS1leGVjdXRpb24tMjAyNi0xMS0xNgDkBhwzu9k93Fr4TQfF6je+0cXwlTQwMDAwIHI0YS1maW5hbC1hdWRpdC0yMDI2LTExLTE2AE/eWodvItsy9CIDjwwlwjqc4BOXNDAwMDAgcjRhLWxhYmVsbGVkLXJlY292ZXJ5LTIwMjYtMTEtMTYAlf/gtJNunsn/ZUE2LHaLfgym1bY0MDAwMCByNGEtcGxhbm5pbmctMjAyNi0xMS0xNgDNiAh8PU/gACs8NriTRtW/JeyLvzQwMDAwIHI0YS1yZWNvdmVyeS1hY2NlcHRhbmNlLTIwMjYtMTEtMTYAXgdU8MZqjlc7V9JVAkPBumflz9w0MDAwMCByNGEtdGVybWluYWwtYXVkaXQtMjAyNi0xMS0xNgDv9wkWvYkFU+ikoL4aHv5f2K+WmDQwMDAwIHI0YmMtYWNjZXB0YW5jZS0yMDI2LTExLTE2AAJ3/9nZDzOl47lsB7fBXMM90EiCNDAwMDAgcjRiYy1leGVjdXRpb24tMjAyNi0xMS0xNgDijrMhJDAcYtpVaxTcDI0S3/znKzQwMDAwIHI0YmMtaW50ZXJpbS0xMDAtMjAyNi0xMS0xNwBRbr87h+owuiNRZWPLMgr9i348gTQwMDAwIHI0YmMtcGxhbm5pbmctMjAyNi0xMS0xNgA8S9+0AqCmfDO0PzCtr7TNTm0IlzQwMDAwIHMxLXBhbmVsLXByZWNoZWNrLTIwMjYtMTEtMTQAOx8e228Xq0ZIM0d5S2Bqy2t25Lo0MDAwMCBzMi1jdXN0b2R5LWV2YWx1YXRpb24tMjAyNi0xMS0xNQAZHMUj0jn0rjxnF64E1Oaa5Kcw8jQwMDAwIHMyLWN1c3RvZHktcGxhbm5pbmctMjAyNi0xMS0xNQA1/YVGKSxRV/VrnFtgewRimMyXBTQwMDAwIHMyLWUxLTIwMjYtMTEtMTQA0PGxdXGw7SJ+7Ugjnq1qyPJ7FAQ0MDAwMCBzMi1lMS1kZWxpdmVyeS0yMDI2LTExLTE0AKWxwNtWAHM/KTn0Ekw0o4UOskquNDAwMDAgczItZW5jb2Rlci1kaWFnbm9zdGljLTIwMjYtMTEtMTUACF1qBAH2GvjJOZltnFnl2Er5ITA0MDAwMCBzMi1nMS1iYW5kMjUtMjAyNi0xMS0xNQBkYjfxoYGvDzWYTs2Rpa2spZBrSDQwMDAwIHMyLWcxLWJhbmQyNS1yZXZpZXctY29tcGFyaXNvbi0yMDI2LTExLTE1APi2WQLfpRGvH/tygZ1J3IYv1AV6NDAwMDAgczItZzEtYmxpbmQxMjAtMjAyNi0xMS0xNAATLuRdY6qJyXbuR82I7m8in7tcvTQwMDAwIHMyLWcxLWJsaW5kMTIwLWRlbGl2ZXJ5LTIwMjYtMTEtMTQA+TrH5wEmGXjnnochy7HynzmpG8A0MDAwMCBzMi1nMS1ibGluZDEyMC1wcmVwYXJhdGlvbi0yMDI2LTExLTE0ALaOaaIQpt+bSmGBcG1r0A0FXuiRNDAwMDAgczItZzEtY2FuZGlkYXRlcy0yMDI2LTExLTE0ALnUEyk0yIDawhZgo0TN+qwTNAsPNDAwMDAgczItZzEtY29tcG9uZW50cy0yMDI2LTExLTE1AHXngFNUM0Tru1pwKEsnDH9DTgH6NDAwMDAgczItZzEtY29weS1jYWxpYnJhdGlvbi0yMDI2LTExLTE0API+WKfbNYZGwzbvn6eFvrUaxcOlNDAwMDAgczItZzEtY29weS1jYWxpYnJhdGlvbi10ZXN0cy0yMDI2LTExLTE0ABDYJ4ZvS28QgcJB+ehgNfNf434CNDAwMDAgczItZzEtY3VzdG9keS1oYW5kb2ZmLTIwMjYtMTEtMTQAPHFzcItOLfbfpLBy2/hCdHRmiAQ0MDAwMCBzMi1nMS1pbnZlbnRvcnktMjAyNi0xMS0xNABfNjN1lly8TrUNSrEhuPjWjAEuHjQwMDAwIHMyLWcxLXJlZ2lzdHJhdGlvbi0yMDI2LTExLTE0AO1DUQC6azQ8qPyUBauvYbU8bujrNDAwMDAgczItZzEtc2NoZWR1bGVyLXNpbXVsYXRpb24tMjAyNi0xMS0xNACqqOrUnyQfLjQKTMJvQ6ItZtCeHTQwMDAwIHMyLW1haW4tZW5naW5lZXJpbmctMjAyNi0xMS0xNQAf3bq9vF5UEBTBJ5fvp1o/eTqdOzQwMDAwIHMyLW1haW4tbWF0cml4LTIwMjYtMTEtMTUAab51ObRkUzGvl0XuOeRcd9QMf6c0MDAwMCBzMi1tYWluLXByZXBhcmF0aW9uLTIwMjYtMTEtMTUAOB1hLKnpf7I0D87dy00P7Tkb2MM0MDAwMCBzMi1tb2RlbC1ub3JtYWxpemF0aW9uLTIwMjYtMTEtMTQAqQqgd9Q9iyjL+gLIq8j8QDeBwSo0MDAwMCBzMi1tb2RlbC1zeW50aGV0aWMtMjAyNi0xMS0xNADEBknPof8CgvudVeGzSQ44i3t0YjQwMDAwIHMyLXBpbG90LWNvbnRpbnVhdGlvbi0yMDI2LTExLTE1AJwhWIYtGeebZPqWr254no4ihbx8NDAwMDAgczItcGlsb3QtaW1wbGVtZW50YXRpb24tMjAyNi0xMS0xNQBaZl8QyV91eWKmR6BFBNHE/YrsNjQwMDAwIHMyLXBpbG90LWlucHV0cy0yMDI2LTExLTE1AMYH4YRWgmn3syHj1Enjel+3XLalNDAwMDAgczItcGlsb3QtcmVjb3ZlcnktY29udGludWF0aW9uLTIwMjYtMTEtMTUAw0w5c42ArKwVMdUrc5nGdeBiyZs0MDAwMCBzMi1waWxvdC1yZWNvdmVyeS1pbnB1dHMtMjAyNi0xMS0xNQBLsSL/QY0e4EClb5uZuGxc4JuX+DQwMDAwIHMyLXBpbG90LXRyYWluaW5nLTIwMjYtMTEtMTUA+wv2AdknTzCDVyewpAJ0um5X63o0MDAwMCBzMi1waWxvdC12Mi0yMDI2LTExLTE1AHQr19W/DaIPVnY9zPtWOtPUtZP3NDAwMDAgczItcGlsb3QtdjItY29udGludWF0aW9uLTIwMjYtMTEtMTUAiTzgq8r+SI/VnE8gbvKnD2VInHU0MDAwMCBzMi1wb3N0aG9jLWRpYWdub3N0aWNzLTIwMjYtMTEtMTUACbtM+GDpYO76JEIFxzffqcn/RY00MDAwMCBzMi1wcmVjaXNpb24tcGlsb3QtaW1wbGVtZW50YXRpb24tMjAyNi0xMS0xNQC0rl/2pCP1VnkakmVxJEdYMEjThzQwMDAwIHMyLXByZXBhcmF0aW9uLWluY2lkZW50LTIwMjYtMTEtMTUAXWIBxvbFoh+V9PqNm3mbad++XHg0MDAwMCBzMi1yMi1tYWluLTIwMjYtMTEtMTUAkPvw7KYMcs+YIseFP+wRo0tDJFg0MDAwMCBzMi1yZWNvdmVyeS1pbXBsZW1lbnRhdGlvbi0yMDI2LTExLTE1ABtA2R3sSZgmsORYU6Dwehil2ETZNDAwMDAgczItcmVjb3ZlcnktcGlsb3Qtc3RhZ2UtcmV2aWV3LTIwMjYtMTEtMTUApy/Gf0ahf9qVUMD+TKksX8XD/780MDAwMCBzMi10ZW4tcGFydC1tZWFzdXJlbWVudC0yMDI2LTExLTE1APZVcWFm1VFwq2NpYzG2LaazVtZTNDAwMDAgczItdGVybWluYWwtZnJlZXplLTIwMjLine truncated
"516ebf3b87ea30ba23516563cb320afd8b7e3c81": "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",
"98f05aa07d1c538c686857fd6402fb14ce457d95": "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 truncated
},
"files": {
"artifacts/r4bc-interim-100-2026-11-17/contract.json": "eb2d68b2f29135ecdc8ccf48165ae95cec9123183e8d03cb04ab9ad1174b3306",
"artifacts/r4bc-interim-100-2026-11-17/models-terminal.json": "0703cd626870a23f84e600a4b09fae656b875a3d7e7f7ce60d6259d81edf132d",
"artifacts/r4bc-interim-100-2026-11-17/interim-freeze.json": "24d564f06f92fc6e2e6649668cfe6847b71ed68452ec5474ba8a31144c42c81f",
"artifacts/r4bc-interim-100-2026-11-17/adapter-freeze.json": "b92265d99089ca6fb6b931ecf4a911394f29275134b05dbdcc1ebd32b62f012a",
"artifacts/r4bc-interim-100-2026-11-17/predictions-freeze.json": "db2ceca68f0eedb94f031c0deb958df234f934fa4534ea444423c4668af3de1e",
"tools/r4bc_interim.py": "8c57850ae0014f7984807d51575c7c9a95efd7b356b9c9b912b9667309919478"
}
}
@@ -0,0 +1,19 @@
# R4bc:最终拟合的条件性基础设施恢复预裁量
已实际读取 [Claude 原串1794847751.356439](https://arcticnimbus.slack.com/archives/C0C78UFKGRM/p1794847751356439),2026-11-17 00:49:11,UTC+8。本记录仅保存后续处置条件;未启动恢复、未修改活跃代码或合同文件。
Claude 原文:
> Pre-ruling to avoid a wait: if a final fit (or an XGBoost A/Q fit, which has no HPO fallback) hits process_count_limit with this signature, MeltAstra may run one labelled infrastructure recovery of that single fit after the run is otherwise terminal, within the same ledger. It is reported as such and is not a retry of a numerical failure. Any other failure type stays a kept failure.
适用条件与边界:
- 仅适用于 final fit,或无 HPO 替补的 XGBoost A/Q fit。
- 必须为同一已描述签名:约1.3秒出现第二进程、`process_count_limit`、无 traceback、子进程身份未明;不能将其他失败泛化为此签名。约1.3秒是原描述,不另造精确数值容差。
- 只有本轮其余任务全部终态后,才允许对符合条件的该单次拟合进行一次明确标记的基础设施恢复。不得提前派发、重复恢复或覆盖原失败。
- 原失败与其费用保留;恢复及原工作共用同一累计账本和既有预算、资源边界,不因本预裁量扩张帽值。后续仍须遵守模型 Git 冻结→预测→预测/评分配置 Git 冻结→评分的既有顺序。
- 明确报告这是经预裁量允许的基础设施恢复,不称为数值失败重试;子进程成因仍未由现有证据确认。其他类型失败原样保留。
job90 是 H1 HPO config10,不属于此恢复范围:原配置NA,从11个成功配置选参、final job92已PASS,Claude确认无需动作。不得重试job90、追补其HPO搜索或改动运行中的固定来源。
原唯一协调器继续依现有合同运行;本记录不代表条件已触发、恢复已执行或恢复已成功。
+4
View File
@@ -17,3 +17,7 @@ Claude原串1794858745.402629允许在原50%/25%训练继续期间,先冻结10
## 真实预测完成
04:06:18已实核独立服务536005/CPU1与实际预测worker536019/CPU0,首H0p 121件PASS;后续35/35预测guard全部PASS,3864份新预测加552历史H0/552EIK,共4968条100%清单已hash冻结。新预测worker257.177371CPU秒;含60预留post账317.177371/14400秒,协调CPU30.789517561另列。原训练队列未改。下一步在此预测清单Git冻结后才评分;尚无分数。
## 评分终态
35预测全部PASS后,预测清单已固定a326870;唯一score服务536613/worker536623于04:11:59实见189件处理中,随后score-complete正常完成、guard PASS。结果见[INTERIM_100固定结果](r4bc-interim-100-results-2026-11-17.md)。552/204完整,post累计346.870387/14400秒,原训练继续。最终全池精确核对仍待未来实际终态。
@@ -0,0 +1,57 @@
# R4(b)/(c) INTERIM_100:固定100%学习池结果
2026-11-17。依Claude1794858745.402629,100%池335任务全部终态后,先冻结35个最终模型,再冻结全部100%预测,使用原固定评分函数。**本报告仅为INTERIM_100,50%/25%训练仍由原队列继续;不是全矩阵终态。**
## 覆盖、主结果与辅助指标
共同552件/204组件,五折外测齐全。各项比较均覆盖完整共同域,无未定义指标;每件counts一次的相对L2,组件内等件、组件间等权。与Gre24随机划分/全tet域RRMSE不同,不混用分母。τ-b越大越好,其余指标越小越好。
| 臂 | 相对L2(主) | Kendall τ-b | 最后位置距离/Rg | 近壁相对L2 |
| --- | ---: | ---: | ---: | ---: |
| EIK | 0.095879482537 | 0.916525652665 | 0.342375611203 | 0.083638948330 |
| H0 | 0.073027071673 | 0.924479045724 | 0.441287552611 | 0.074931309083 |
| H0p | 0.073022990543 | 0.924674448296 | 0.418977284803 | 0.074963075616 |
| H1 | 0.074262578885 | 0.923068425108 | 0.400495063672 | 0.076021959281 |
| H2 | 0.072051346543 | 0.925562040015 | 0.424134938005 | 0.074215496368 |
| H3 | 0.070024363499 | 0.927193927492 | 0.378983996603 | 0.071904839589 |
| H4 | 0.070426230068 | 0.927592410717 | 0.399624182857 | 0.072322301156 |
| A | 0.070830540821 | 0.926649719537 | 0.376713807712 | 0.072787154312 |
| Q | 0.072781835831 | 0.923894847343 | 0.434439880835 | 0.075026127258 |
H0为历史固定预测;H0p即H0′,在共同名单重新按原12配置HPO训练。H1删t_set输入,H2换整套几何构造,H3在同查询上用单位权,H4扩展到全节点单位权,A换为固定XGB,Q为XGB查询层级权变体。A与H4的差同时包含学习器、浮点接口与HPO对固定参数的训练政策差异。
## R4(b)与逐步差值
以下Δ均为**前臂误差−后臂误差的绝对差**,正值表示后臂误差较低,不是相对提升百分比。95%区间为204组件配对bootstrap 2000次、seed2026111408的描述性percentile区间,条件于固定模型/折;无新增通过门。
| 比较 | Δ | 95%区间 |
| --- | ---: | --- |
| EIK->A | 0.025048941717 | [0.022195926818, 0.027818408847] |
| EIK->Q | 0.023097646706 | [0.020387248607, 0.025891362837] |
| Q->A | 0.001951295010 | [0.001223714800, 0.002679049562] |
| H0->H0p | 0.000004081129 | [-0.000232486957, 0.000236333961] |
| H0p->H1 | -0.001239588342 | [-0.002015916781, -0.000492297742] |
| H1->H2 | 0.002211232342 | [0.001466705021, 0.002872386974] |
| H2->H3 | 0.002026983044 | [0.001414488614, 0.002665400914] |
| H3->H4 | -0.000401866569 | [-0.000940178294, 0.000165025748] |
| H4->A | -0.000404310752 | [-0.000744109302, -0.000084007680] |
在这一固定100%池上,删t_set输入使误差增加;换整套几何构造、在相同查询上改单位权各自降低误差。再扩大全节点的差值区间跨零;末步换为A略增误差,其区间未跨零。H3在本表中点值最低,但本次没有用这些测试结果重新选择模型、调整后续训练或宣称跨学习池样本效率结论。A相对Q误差较低,EIK与学习模型差异明显。H0→H0′区间跨零,包含名单/HPO响应,不作为五步之一。
## 原始失败与执行边界
100%池335终态=334PASS+原job90 HPO process_count_limit;35个final均PASS。job90保持原失败,H1该格从11成功配置选config3,其final已PASS,无重试。额外进程身份未明,不称数值失败;其他池仍运行,本报告不覆盖其未来失败或恢复。
35个预测guard全部PASS,生成3864份新预测;复用552个H0和552个EIK,清单4968条。评分guard PASS,4968逐件/臂评分,四项指标均无未定义值。仅读取原缓存测试标签,VTU新解码0、新拟合0。主/链/每条边均552件/204组件;50%/25%的空摘要是原评分函数的占位输出,明确不作为结果。
固定阶段:模型/代码/原回执提交`1f5ce913bcd424290162ceb8d94e6fa4cff98a12`;评分前完整新预测及清单提交`a326870216f34e1e38d90319228e92846838675b`,预测清单SHA256 `db2ceca68f0eedb94f031c0deb958df234f934fa4534ea444423c4668af3de1e`。原合同与计算代码均未修改,interim入口只改变范围门和编排,没有伪造整轮终态。
预测worker257.177371CPU秒、评分29.693016CPU秒;加60秒工程/元数据预留,**累计post346.870387/14400秒,剩余14053.129613秒**。原训练费用仍另由fit账记录,最终post必须累计本次费用,不能重开4h预算。两阶段协调CPU分别30.789517561、3.413776秒,单列。
## 最终结果一致性与解释限制
最终全池评分必须精确复现本INTERIM_100:相同原合同、35模型SHA、100%预测数组、逐件分数、汇总及bootstrap。检查工具`tools/r4bc_interim_compare.py`已发布准备,零容差,差异输出DEFECT;两项合成验证通过,包括最小浮点差异被拒绝。最终全池尚未完成,**本报告不声称最终一致性已经验证**。
原数据覆盖、共同名单排除、已暴露的固定公共样本、各步含训练政策响应等限制保留。本结果不能关闭独立类型/拓扑/连通假说,不能单凭100%池建立样本效率或新工艺泛化结论;由Claude统筹阶段解释。原50%/25%队列不因本结果改变。
证据入口:[INTERIM_100机器结果](../artifacts/r4bc-interim-100-2026-11-17/INTERIM_100.json)、[原评分摘要](../artifacts/r4bc-interim-100-2026-11-17/scores/summary.json)、[评分guard](../artifacts/r4bc-interim-100-2026-11-17/score-guard/worker.receipt.json)、[累计后处理账](../artifacts/r4bc-interim-100-2026-11-17/score-complete.json)。