{
"candidate_id": "ATT-0086",
"reviewer": "D",
"rubric": "PAPER-RUBRIC-1.0",
"review_status": "frozen",
"identity": {
"candidate_id": "ATT-0086",
"title": "CrossLinear: Plug-and-Play Cross-Correlation Embedding for Time Series Forecasting with Exogenous Variables",
"authors": [
"Zhou, Pengfei",
"Liu, Yunlong",
"Liang, Junli",
"Song, Qi",
"Li, Xiangyang"
],
"arxiv_id": "2505.23116",
"arxiv_version": "v1",
"first_arxiv_date": "2025/05/29",
"latest_arxiv_date": "2025/05/29",
"source_urls": [
"https://doi.org/10.1145/3711896.3736899"
],
"publication_status": "KDD 2025, DOI 10.1145/3711896.3736899 verified",
"full_text_path": "/workspace/scratch/ac01e91af073/candidate_recovery/priority_packet_20261108/papers/ATT-0086/ATT-0086_paper.pdf",
"metadata_source_path": "/workspace/scratch/ac01e91af073/candidate_recovery/priority_packet_20261108/papers/ATT-0086/ATT-0086_arxiv.html",
"full_text_sha256": "0235fd4cd7a51ae05b313532abdc7d0fe4dc86ac23bc29c1105f5e3cccb814f8",
"review_state": "ready_for_five_new_independent_reviews",
"completed_new_reviews": 0,
"source_versions_not_asserted_identical": true
},
"analysis_date": "2026-11-08",
"reading_record": "2026-11-08 UTC：先读历史条目与身份，再逐页阅读全文物理页1—13，包括p10参考文献和p11—13全部附录；逐表核查核心主张、退化结果和重参数化推导，视觉检查p4、8、9、13的公式、遮挡表、热图与示例。大表未逐单元重算。",
"freeze_time_utc": "2026-11-08T02:59:05.628285+00:00",
"review_type": "independent_full_text_review",
"independence": "未读取其他新评；先读取旧历史条目，再检查本版本全文和附录。",
"legacy": {
"full_review_available": false,
"counts_as_review": false,
"source_sha256": "5d337d3b19381048f4211cdc36d6d21050166cddfbddc7a1a942119d33a63270",
"records": [
{
"source_record_id": "S5-086",
"source_id": "S5",
"source_library_file_id": "libfile_849759d66a808191813dbe9bd1605bb4",
"source_filename": "已粘贴的文本(5).txt",
"source_line": 28,
"source_number": 86,
"title": "CrossLinear: Plug-and-Play Cross-Correlation Embedding for Time Series Forecasting with Exogenous Variables",
"source_url": "https://doi.org/10.1145/3711896.3736899",
"raw_line": "| 12 | [P086 CrossLinear: Plug-and-Play Cross-Correlation Embedding for Time Series Forecasting with Exogenous Variables](https://doi.org/10.1145/3711896.3736899) | 59.2 | 100% | 55% | 53.7–65.0 | 18 | arXiv:2505.23116v1；本地 PDF；ACM 论文集版本未单独读取 | **重要问题**：主表及消融未报告重复运行次数、随机种子或误差区间；例如 ECL 多对一平均 MSE 的差距为 0.004，稳定性未知。；
**重要问题**：附录 B 给出 Sum 到 Cross Only 的卷积重参数化；单靠表 4 的结果无法区分参数化、初始化、训练动态和约束作用，也缺权重物理含义的独立验证。；
**重要问题**：插件在部分设置下出现退化：表 11 中 Traffic 96 步的 DLinear MSE 从 0.268 升至 0.338；ECL 的 DLinear 平均 MAE 从 0.457 升至 0.463。 | [评审](论文/P086/评审.md) · [评分数据](论文/P086/评审.json) · 无单独复核文件 |",
"raw_cells": [
"12",
"[P086 CrossLinear: Plug-and-Play Cross-Correlation Embedding for Time Series Forecasting with Exogenous Variables](https://doi.org/10.1145/3711896.3736899)",
"59.2",
"100%",
"55%",
"53.7–65.0",
"18",
"arXiv:2505.23116v1；本地 PDF；ACM 论文集版本未单独读取",
"**重要问题**：主表及消融未报告重复运行次数、随机种子或误差区间；例如 ECL 多对一平均 MSE 的差距为 0.004，稳定性未知。；
**重要问题**：附录 B 给出 Sum 到 Cross Only 的卷积重参数化；单靠表 4 的结果无法区分参数化、初始化、训练动态和约束作用，也缺权重物理含义的独立验证。；
**重要问题**：插件在部分设置下出现退化：表 11 中 Traffic 96 步的 DLinear MSE 从 0.268 升至 0.338；ECL 的 DLinear 平均 MAE 从 0.457 升至 0.463。",
"[评审](论文/P086/评审.md) · [评分数据](论文/P086/评审.json) · 无单独复核文件"
],
"normalized_title": "crosslinearplugandplaycrosscorrelationembeddingfortimeseriesforecastingwithexogenousvariables",
"normalized_url": "doi.org/10.1145/3711896.3736899",
"record_type": "legacy_scored",
"legacy_rank": 12,
"legacy_score": "59.2",
"legacy_coverage": "100%",
"legacy_tentative_share": "55%",
"legacy_range": "53.7–65.0",
"citation_verbatim": "18",
"legacy_review_version": "arXiv:2505.23116v1；本地 PDF；ACM 论文集版本未单独读取",
"legacy_risks_verbatim": "**重要问题**：主表及消融未报告重复运行次数、随机种子或误差区间；例如 ECL 多对一平均 MSE 的差距为 0.004，稳定性未知。；
**重要问题**：附录 B 给出 Sum 到 Cross Only 的卷积重参数化；单靠表 4 的结果无法区分参数化、初始化、训练动态和约束作用，也缺权重物理含义的独立验证。；
**重要问题**：插件在部分设置下出现退化：表 11 中 Traffic 96 步的 DLinear MSE 从 0.268 升至 0.338；ECL 的 DLinear 平均 MAE 从 0.457 升至 0.463。",
"legacy_material_links_verbatim": "[评审](论文/P086/评审.md) · [评分数据](论文/P086/评审.json) · 无单独复核文件",
"counts_as_independent_review": false
}
]
},
"claims": [
"用残差跨通道卷积构建轻量历史协变量融合，再接分块线性预测头。",
"多对一与多对多十二数据集比较，并在五骨干中测试可插拔性。",
"作者证明Sum与Cross Only可作卷积重参数化，收益需按优化和归纳偏置解释。"
],
"review_body": "CrossLinear面对的是一个有实际意义的设定：只预测目标序列，但允许使用其他变量的历史，避免让多变量模型同时承担不需要的预测任务。它用一个跨变量的一维卷积生成单条嵌入，再与目标序列作残差加权，接分块线性投影和全局预测头。设计简洁，提供了比复杂注意力更低容量的变量融合方案。本文最可信的贡献是这种轻量参数化在多组公开任务中的工程效果与可插拔性，而不是发现新的互相关统计量。卷积权重由预测损失学习，模型没有显式计算或约束统计意义的交叉相关函数；时间不变来自共享权重的结构约束。“只保留直接依赖、排除间接依赖”也只应理解为网络连接形式，不能当作数据因果关系已经识别。\n\n关键的概念边界在作者自己的附录证明里。第十一至十三页把加权残差Sum改写为新的卷积核，说明其与Cross Only有相同的线性表达能力。这一说明是优点，不是论文隐瞒的漏洞；但也意味着表四Sum优于Cross Only不能证明它引入了独有的相关信息。初始化、残差偏置、隐式正则化和优化路径都可能解释差异。第九页确实讨论初始化靠近一的优势，因此更稳妥的结论是该参数化有利于优化。当前没有匹配有效初始卷积核和正则项的对照证据，机制归因应保留。位置嵌入接在线性投影与展平线性头之间，在固定长度下其输入无关贡献可被写成一个输出偏置；它改善结果并不自动证明模型新获得了顺序感知能力，展平头本来就区分位置。\n\n主实验较为扎实地覆盖了十二个数据集、长短预测及多对一和多对多两种任务，TimeXer是直接相关且有竞争力的对手，此外包含TiDE、iTransformer、SparseTSF等。多对一ECL平均MSE为零点三二三，相比TimeXer零点三二七略好；Traffic为零点一五二对零点一五六。如此小的差距在没有重复运行统计时不能写成稳定胜出。较长跨度也有明确失败，例如ECL七百二十步零点四零三不及TimeXer零点三六五。多对多Traffic平均MSE零点四八三，明显落后iTransformer零点四二八，说明把每个变量都视为目标后，简单共享预测器并不总占优。电价任务多对一平均MSE优于TimeXer，而MAE零点二六六略差于零点二六五，必须保留指标差异。\n\n公平性方面，论文明确数据划分、固定历史长度、预测跨度与十轮训练，附录还给学习率、批量和卷积设置，并说明基线遵循原论文或官方配置。统一十轮使预算较清楚，但不同模型收敛速度不同，不能保证各架构都到达可比的最优训练状态。学习率和批量区间之外，选择规则、每个基线实际搜索次数及全部最终配置仍不充分。作为外生预测论文，ARIMAX、正则化线性滞后回归和文中引用的NBEATSx没有进入主比较，因而尚不清楚它相对传统外生建模强基线的增量。数据中的其他测量通道在此是预测协变量，不等同于经过统计外生性验证的变量；公式只使用历史值，也不应推断论文测试了未来已知协变量的价值。\n\n消融与插件实验是值得保留的证据。Endo Only、Cross Only、Concat、位置嵌入和超参数曲线覆盖核心组件；把模块插入五种模型进一步展示可复用性。不过不是逐项都改善：附录第十二页Traffic九十六步DLinear的MSE从零点二六八增至零点三三八，ECL的DLinear平均MAE由零点四五七增至零点四六三，个别PatchTST任务也退化。正文部分措辞先说每次加入都有改善，随后又改为大多数改善，前者应收窄。权重热图与若干气象关系看起来相符，只是定性一致性，不能排除共线性、尺度变换或重新参数化造成的等价权重；三点核求和还会抹掉不同滞后位置的符号抵消。因此“物理依赖恢复”仍未获得独立真值或干预证据。\n\n缺失值部分比完全不测鲁棒性更好，但不能直接认定优于其他模型。实验只列本模型在随机置零或随机数替换下的误差，没有同样破坏条件的强基线；连续缺失块、非随机缺失和错误协变量都不在当前覆盖内。尤其当目标历史全部缺失时，RevIN的均值方差如何计算、遮挡发生于归一化之前还是之后、训练和测试是否都遮挡，这些步骤没有完整交代。它们会影响是否仍保留原目标历史统计信息，当前只能标为未知而非指控泄漏。表六将NP也标成九十六到九十六，而正文电价设置是一百六十八到二十四，具体协议尚需区分。\n\n资源结论也应限定范围。固定通道数、预测长度和宽度时，随输入长度线性增长是合理的；但多对多卷积权重形状明确是变量数乘变量数乘三，对通道数不是线性。全局头还随输出长度和嵌入宽度增长，不能由一个关于历史长度的大O符号推出“近乎零成本”或任意高维可扩展。ECL同批量每步训练图显示良好精度速度折中，但没有完整显存、端到端耗时和多硬件测量。代码同时给GitHub和Zenodo归档入口，提高可获得性；本次未打开运行仓库或核验归档一致性，未执行训练、推理和复现。综合来看，这是有用的轻量外生融合基线，主表与插件实验支持继续研究它，但新颖性、因果式依赖解释、稳定领先与通用部署叙述都需要保持克制。",
"review_chinese_character_count": 1651,
"dimensions": [
{
"dimension": "A1",
"name": "问题重要性",
"score": 85,
"status": "rated",
"confidence": "high",
"reason": "只预测目标且利用历史协变量的效率问题实用且常见",
"evidence_pages": "1—3",
"judgment_range": null
},
{
"dimension": "A2",
"name": "原创性",
"score": 60,
"status": "tentative",
"confidence": "mid",
"reason": "轻量残差卷积融合有具体价值，基础算子与表达空间并非新发明",
"evidence_pages": "3—4,11—13",
"judgment_range": [
50,
70
]
},
{
"dimension": "A3",
"name": "实质推进程度",
"score": 70,
"status": "rated",
"confidence": "mid",
"reason": "十二数据集双任务与插件实验形成工程推进，小差距稳定性未知",
"evidence_pages": "5—8,12",
"judgment_range": null
},
{
"dimension": "A4",
"name": "知识与解释增量",
"score": 55,
"status": "rated",
"confidence": "mid",
"reason": "明确重参数化等价性有价值，直接依赖和权重解释缺独立验证",
"evidence_pages": "9,11—13",
"judgment_range": null
},
{
"dimension": "B1",
"name": "问题定义与评价有效性",
"score": 75,
"status": "rated",
"confidence": "mid",
"reason": "历史协变量与预测目标明确，缺失协议和NP表头口径未清",
"evidence_pages": "3—5,8",
"judgment_range": null
},
{
"dimension": "B2",
"name": "技术与论证正确性",
"score": 75,
"status": "rated",
"confidence": "mid",
"reason": "主模型和卷积重参数化技术上可理解，复杂度需限定固定变量维度",
"evidence_pages": "4,9,11—13",
"judgment_range": null
},
{
"dimension": "B3",
"name": "数据与评测流程完整性",
"score": 60,
"status": "tentative",
"confidence": "mid",
"reason": "数据划分训练轮数可追踪，选择规则、重复与缺失归一化流程未全说明",
"evidence_pages": "5,8,11",
"judgment_range": [
50,
75
]
},
{
"dimension": "B4",
"name": "比较对象与比较公平性",
"score": 65,
"status": "rated",
"confidence": "mid",
"reason": "TimeXer等相关强基线齐全，未含传统外生回归和NBEATSx，十轮不等于同等收敛",
"evidence_pages": "5—7,10—11",
"judgment_range": null
},
{
"dimension": "C1",
"name": "核心主张的证据充分度",
"score": 70,
"status": "rated",
"confidence": "mid",
"reason": "主表与插件覆盖广，普遍领先和机制解释超过实证",
"evidence_pages": "6—9,12",
"judgment_range": null
},
{
"dimension": "C2",
"name": "机制归因与竞争解释区分度",
"score": 55,
"status": "rated",
"confidence": "high",
"reason": "四配置与位置消融存在，但表达等价使初始化优化成为强竞争解释",
"evidence_pages": "7,9,11—13",
"judgment_range": null
},
{
"dimension": "C3",
"name": "结果稳定性与不确定性处理",
"score": 30,
"status": "rated",
"confidence": "high",
"reason": "没有种子、重复次数或误差区间，小差距无法判断稳定性",
"evidence_pages": "6—8,11—12",
"judgment_range": null
},
{
"dimension": "C4",
"name": "适用范围与边界验证",
"score": 70,
"status": "rated",
"confidence": "mid",
"reason": "长短预测、双范式、随机缺失均覆盖，但分布漂移与真实缺失仍未验证",
"evidence_pages": "5—8",
"judgment_range": null
},
{
"dimension": "D1",
"name": "过程透明度",
"score": 65,
"status": "rated",
"confidence": "mid",
"reason": "优化器范围设备和轮数公开，最终配置及完整遮挡流程不足",
"evidence_pages": "8,11",
"judgment_range": null
},
{
"dimension": "D2",
"name": "研究资源可获得性",
"score": 75,
"status": "tentative",
"confidence": "mid",
"reason": "给GitHub和Zenodo归档，未核实一致性与可运行性",
"evidence_pages": "1",
"judgment_range": [
60,
85
]
},
{
"dimension": "D3",
"name": "主张克制与报告一致性",
"score": 55,
"status": "rated",
"confidence": "high",
"reason": "普遍改善、直接依赖和近乎零成本过强，表格保留了退化结果",
"evidence_pages": "1—2,8—9,12",
"judgment_range": null
},
{
"dimension": "D4",
"name": "表达清晰度",
"score": 75,
"status": "rated",
"confidence": "mid",
"reason": "结构公式和附录较完整，维度文字、表头与条件性叙述有瑕疵",
"evidence_pages": "3—9,11—13",
"judgment_range": null
},
{
"dimension": "E1",
"name": "复杂度合理性",
"score": 80,
"status": "rated",
"confidence": "mid",
"reason": "骨干简洁且残差归纳偏置明确，但多对多跨通道卷积二次增长",
"evidence_pages": "4,9",
"judgment_range": null
},
{
"dimension": "E2",
"name": "资源效率与成本收益",
"score": 65,
"status": "tentative",
"confidence": "mid",
"reason": "同批量单数据训练图有优势，缺完整显存推理和端到端成本",
"evidence_pages": "9,11",
"judgment_range": [
50,
80
]
},
{
"dimension": "E3",
"name": "可迁移与可复用程度",
"score": 80,
"status": "rated",
"confidence": "high",
"reason": "模块在五种骨干实测可移植，存在需保留的个别负迁移",
"evidence_pages": "8,12",
"judgment_range": null
},
{
"dimension": "E4",
"name": "后续研究启发性",
"score": 80,
"status": "rated",
"confidence": "mid",
"reason": "有助研究外生变量价值、重参数化和优化，而非仅换复杂骨干",
"evidence_pages": "7—9,11—13",
"judgment_range": null
},
{
"dimension": "F1",
"name": "与当前问题的相关性",
"score": null,
"status": "unknown",
"confidence": "low",
"reason": "未知用户当前研究问题",
"evidence_pages": "用户背景未提供",
"judgment_range": null
},
{
"dimension": "F2",
"name": "相对于已有知识的学习收益",
"score": null,
"status": "unknown",
"confidence": "low",
"reason": "未知用户已掌握知识",
"evidence_pages": "用户背景未提供",
"judgment_range": null
},
{
"dimension": "F3",
"name": "当前使用可行性",
"score": null,
"status": "unknown",
"confidence": "low",
"reason": "未知用户设备、时间和数据约束",
"evidence_pages": "用户背景未提供",
"judgment_range": null
},
{
"dimension": "F4",
"name": "当前决策信息价值",
"score": null,
"status": "unknown",
"confidence": "low",
"reason": "未知用户具体选择与机会成本",
"evidence_pages": "用户背景未提供",
"judgment_range": null
}
],
"missing_and_limits": [
"旧完整评语缺失，仅有历史行记录，未作新评票。",
"全文未给随机种子、重复次数、方差和置信区间。",
"缺失值遮挡与RevIN统计量计算顺序不清，NP表头与主体协议存在歧义。",
"没有执行训练、推理、代码测试或复现；GitHub/Zenodo一致性未核验。",
"F组个人背景和决策需求未知。"
],
"experiments_executed": false,
"aggregate_score": null
}