{ "rubric_version": "PAPER-RUBRIC-1.0", "paper_id": "ATT-0097", "paper_title": "Adaptive Multi-Scale Decomposition Framework for Time Series Forecasting", "reviewer": "B", "analysis_mode": "standard", "paper_version": "2406.03751v2", "version_date": "2025/04/15", "first_public_date": "2024/06/06", "assessment_date": "2026-11-08", "reading_started_at": "2026-11-08T02:45:31.018504+00:00", "reading_completed_at": "2026-11-08T02:47:39.067212+00:00", "read_physical_pages": [ 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14 ], "visual_checks_physical_pages": [ 4, 6, 7, 10, 11, 12, 13, 14 ], "pdf_sha256": "02aad9b5b876e3b53d6c34114d6e658eb557898d57f20402baf512b3e264b05e", "legacy_review_read_first": true, "legacy_full_review_available": false, "legacy_status": "仅有历史清单行或风险摘要；未提供完整旧评，不能声称复读完整旧评。", "legacy_records": [ { "source_record_id": "S5-097", "source_id": "S5", "source_library_file_id": "libfile_849759d66a808191813dbe9bd1605bb4", "source_filename": "已粘贴的文本(5).txt", "source_line": 29, "source_number": 97, "title": "Adaptive Multi-Scale Decomposition Framework for Time Series Forecasting", "source_url": "https://ojs.aaai.org/index.php/AAAI/article/view/33908", "raw_line": "| 13 | [P097 Adaptive Multi-Scale Decomposition Framework for Time Series Forecasting](https://ojs.aaai.org/index.php/AAAI/article/view/33908) | 59.2 | 100% | 35% | 55.5–63.5 | 74 | AAAI 2025 会议论文；补充理论与结果为 arXiv:2406.03751v2（2025-04-15） | **重要问题**：arXiv v2 附录 A 式 (13)→(14) 把分母由 a-b 变为 d(a-b) 时遗漏 d；式 (21)→(22) 的不等式也忽略不同列权重。；
**重要问题**：表 1 的 ETTh2/720 中 AMD MSE 0.402，TimeMixer 0.342；长期预测表现取决于数据与预测长度。；
**重要问题**：基线各模型调参预算和最佳输入长度选择程序未在已检查文字中完整交代；影响相对效果的可比性。 | [评审](论文/P097/评审.md) · [评分数据](论文/P097/评审.json) · 无单独复核文件 |", "raw_cells": [ "13", "[P097 Adaptive Multi-Scale Decomposition Framework for Time Series Forecasting](https://ojs.aaai.org/index.php/AAAI/article/view/33908)", "59.2", "100%", "35%", "55.5–63.5", "74", "AAAI 2025 会议论文；补充理论与结果为 arXiv:2406.03751v2（2025-04-15）", "**重要问题**：arXiv v2 附录 A 式 (13)→(14) 把分母由 a-b 变为 d(a-b) 时遗漏 d；式 (21)→(22) 的不等式也忽略不同列权重。；
**重要问题**：表 1 的 ETTh2/720 中 AMD MSE 0.402，TimeMixer 0.342；长期预测表现取决于数据与预测长度。；
**重要问题**：基线各模型调参预算和最佳输入长度选择程序未在已检查文字中完整交代；影响相对效果的可比性。", "[评审](论文/P097/评审.md) · [评分数据](论文/P097/评审.json) · 无单独复核文件" ], "normalized_title": "adaptivemultiscaledecompositionframeworkfortimeseriesforecasting", "normalized_url": "ojs.aaai.org/index.php/aaai/article/view/33908", "record_type": "legacy_scored", "legacy_rank": 13, "legacy_score": "59.2", "legacy_coverage": "100%", "legacy_tentative_share": "35%", "legacy_range": "55.5–63.5", "citation_verbatim": "74", "legacy_review_version": "AAAI 2025 会议论文；补充理论与结果为 arXiv:2406.03751v2（2025-04-15）", "legacy_risks_verbatim": "**重要问题**：arXiv v2 附录 A 式 (13)→(14) 把分母由 a-b 变为 d(a-b) 时遗漏 d；式 (21)→(22) 的不等式也忽略不同列权重。；
**重要问题**：表 1 的 ETTh2/720 中 AMD MSE 0.402，TimeMixer 0.342；长期预测表现取决于数据与预测长度。；
**重要问题**：基线各模型调参预算和最佳输入长度选择程序未在已检查文字中完整交代；影响相对效果的可比性。", "legacy_material_links_verbatim": "[评审](论文/P097/评审.md) · [评分数据](论文/P097/评审.json) · 无单独复核文件", "counts_as_independent_review": false } ], "new_peer_reviews_read": false, "review_text": "先读历史风险摘要，再独立阅读2406.03751v2全部十四页及附录。旧记录提示两个证明步骤、ETTh2长预测和基线选择问题，本次把这些作为待核线索，未继承旧总分，也未得到完整旧评。论文把平均下采样的多尺度残差混合、带通道缩放的顺序块交互以及稠密多预测器门控结合，动机是不同尺度对未来的贡献会变化。相较固定平均，逐预测位置产生权重是有用的工程思想；但已有TimeMixer、混合专家和分块建模背景意味着原创性主要在组合与适配，不能仅凭命名认定全新范式。\n\n历史理论风险在原PDF中成立。第10页式13到14把分母从a-b换成d(a-b)，却保留外面的1/d，漏掉一个d。取f(j)=j、d=2，池化后f1(a)=2a-0.5，相邻粗索引斜率为2，不能由原K=1得到文中仍为1的界。这不否定存在某个更大Lipschitz常数，却否定当前证明宣称的常数保持。式21到22则把不同列权重之差吸收到函数差中；令粗序列恒为1，而两个权重列不同，左侧可非零，右侧函数差为零。固定矩阵并不等于所有时间列相同，所以归纳缺少关键项。\n\n证明后半还有独立范围问题：第10页直接令g(L+t)=g(P+1+t)，未给L=P+1或等价条件；预测构造使用g(P+1)，当非周期情况下P趋向无穷，它不再是有限历史内可用的索引。即使修改为正确的有界误差命题，有限预测区间上有界并不自动说明多尺度比单尺度误差更小。理论不含学习、泛化误差或与基线的界比较，不能为模型优势提供已完成的保证。本次暂停理论保证这一主张，保留实验结果作为独立经验材料，不把证明失败推广成所有预测结果无效。\n\n方法说明也存在影响复现的歧义。第4页式10的门控变换使用α，第12页表6多数数据给α=0。若按印刷符号直接代入，非TopK项为0、TopK项为-1，后续Softmax反而降低所谓主导项权重；表6也可能误把通道缩放β标成α，当前未读实现，不能判定实际训练做反了。第6页可视化解释TP16与TP5，而第7、12页称所有实验八个预测器，这一例图是否另用配置没有说明。摘要所说自相关在主体公式中也没有独立、明确的自相关计算，前馈门控与自相关机制不能不加区别地等同。\n\n实验优点是真实长短期任务较广，声明所有基线重跑并搜索六种输入长度，而且报告五次重复，第14页列出六类数据集的均值和标准差。这里不能写成“完全没有重复实验”。但基线配对波动、各超参预算、最优长度依据验证集还是测试集选择没有完整交代，选择偏差风险应标未知而非直接判泄漏。主要表格显示很强但非一致领先：ETTh2的720步MSE为0.402，TimeMixer为0.342；Exchange的720步DLinear为0.643，本作为0.750；Traffic的96步两项误差也不及TimeMixer。短期表中四组MSE第一，但若干MAE不第一；且缺TimeMixer等强基线在短期中的对应比较。摘要与结论的持续最优措辞超过了这些边界。\n\n消融覆盖去除多尺度、去除交互、平均门控、随机顺序、稀疏专家及均衡损失，是较完整的部件必要性检查。不过仅在Weather与ECL两个集合，删除多预测器同时改变容量，随机打乱属于严重输入干扰，不能完全排除容量、正则化与训练难度解释。表8中d=1没有真正降采样，却随层数也能改善，提示更深映射本身贡献不可忽略；这削弱“全部收益来自多尺度”的独占归因。权重图只是选择性可视化，尚无尺度真值证明专家已分离成真实物理因素。与注意力对比时把自注意力一概说成丢失顺序，也忽略带位置信息的实际模型。\n\n效率只在Weather、输入512和预测96的一种配置展示。图5的八专家模型约1349MB、17毫秒，优于图中较重Transformer，但DLinear约999MB、7毫秒，因而不能宣称对全部MLP同时占优。给DLinear或MTS-Mixers加模块的表4有局部退化，且没有相应资源表，“所有模型均增强且保持同样资源”的说法证据不足。官方仓库readme可读，给Python3.10、数据入口和脚本位置；没有执行代码或确认完整运行链，入口可用不等于复现完成。\n\n总体而言，适合作为多尺度门控预测的经验候选，理论保证与普遍最优需要单独撤回或收窄。尚缺的是门控符号准确映射、同容量收益归因和选择程序记录，而不是要求用户补做一组实验。一个有信息价值的研究问题是专家权重变化究竟代表尺度切换还是普通集成正则化；若无需真实尺度区分也有相同增益，解释性主张就会变弱。优先亲读第4页式7—10与第12页表6的对应，再读第10页证明，最后以第5页失败设置校正总体印象。", "review_hanzi_count": 1383, "claims": [ { "id": "CL01", "claim": "多尺度门控在若干预测任务有效", "status": "supported_with_scope", "pages": [ 5, 6, 14 ] }, { "id": "CL02", "claim": "当前定理证明提供有效理论保证", "status": "suspended", "pages": [ 10 ] }, { "id": "CL03", "claim": "对各数据和长度持续最优", "status": "contradicted_by_tables", "pages": [ 5, 6 ] }, { "id": "CL04", "claim": "专家权重识别真实尺度且提升独立于容量", "status": "not_identified", "pages": [ 6, 14 ] } ], "scores": [ { "dimension_id": "A1", "dimension_name": "问题重要性", "score": 80, "status": "rated", "confidence": "高", "judgment_range": null, "rationale": "多尺度与通道相关性随任务变化是重要预测问题。", "evidence_pages": [ 1, 2 ], "missing_information": "", "source_ids": [ "S01" ] }, { "dimension_id": "A2", "dimension_name": "原创性", "score": 60, "status": "tentative", "confidence": "中", "judgment_range": [ 50, 70 ], "rationale": "多尺度混合与门控组合有一定设计价值，先行工作全文未独立覆盖。", "evidence_pages": [ 2, 3, 4 ], "missing_information": "见理由与材料边界", "source_ids": [ "S01" ] }, { "dimension_id": "A3", "dimension_name": "实质推进程度", "score": 65, "status": "rated", "confidence": "高", "judgment_range": null, "rationale": "多任务有效经验候选，优势不均匀且理论支撑失败。", "evidence_pages": [ 5, 6, 10 ], "missing_information": "", "source_ids": [ "S01" ] }, { "dimension_id": "A4", "dimension_name": "知识与解释增量", "score": 50, "status": "rated", "confidence": "高", "judgment_range": null, "rationale": "部件经验有信息，但尺度解释与理论错误削弱知识增量。", "evidence_pages": [ 6, 7, 10 ], "missing_information": "", "source_ids": [ "S01" ] }, { "dimension_id": "B1", "dimension_name": "问题定义与评价有效性", "score": 75, "status": "rated", "confidence": "高", "judgment_range": null, "rationale": "长短预测与常用误差指标清楚，输入长度选择集未知。", "evidence_pages": [ 5, 6, 11 ], "missing_information": "", "source_ids": [ "S01" ] }, { "dimension_id": "B2", "dimension_name": "技术与论证正确性", "score": 30, "status": "rated", "confidence": "高", "judgment_range": null, "rationale": "证明漏d、忽略权重列差，另有周期索引假设缺失；门控符号歧义。", "evidence_pages": [ 4, 10, 12 ], "missing_information": "", "source_ids": [ "S01" ] }, { "dimension_id": "B3", "dimension_name": "数据与评测流程完整性", "score": 65, "status": "rated", "confidence": "高", "judgment_range": null, "rationale": "提供划分和超参，完整验证选择与归一化流程仍不足。", "evidence_pages": [ 5, 11, 12 ], "missing_information": "", "source_ids": [ "S01" ] }, { "dimension_id": "B4", "dimension_name": "比较对象与比较公平性", "score": 65, "status": "rated", "confidence": "高", "judgment_range": null, "rationale": "重跑且搜索基线上下文值得肯定，调参预算及短期强基线不完整。", "evidence_pages": [ 5, 6, 12 ], "missing_information": "", "source_ids": [ "S01" ] }, { "dimension_id": "C1", "dimension_name": "核心主张的证据充分度", "score": 60, "status": "rated", "confidence": "高", "judgment_range": null, "rationale": "支持部分任务有效，不支持持续最优及现有理论保证。", "evidence_pages": [ 5, 6, 10 ], "missing_information": "", "source_ids": [ "S01" ] }, { "dimension_id": "C2", "dimension_name": "机制归因与竞争解释区分度", "score": 55, "status": "rated", "confidence": "高", "judgment_range": null, "rationale": "多项消融有效，d=1改善提示容量/深度竞争解释仍在。", "evidence_pages": [ 6, 14 ], "missing_information": "", "source_ids": [ "S01" ] }, { "dimension_id": "C3", "dimension_name": "结果稳定性与不确定性处理", "score": 70, "status": "rated", "confidence": "高", "judgment_range": null, "rationale": "五种子和六组标准差提供稳定性，基线配对及消融区间不足。", "evidence_pages": [ 5, 12, 14 ], "missing_information": "", "source_ids": [ "S01" ] }, { "dimension_id": "C4", "dimension_name": "适用范围与边界验证", "score": 65, "status": "rated", "confidence": "高", "judgment_range": null, "rationale": "广泛长短任务，含反例，机制消融仅两组。", "evidence_pages": [ 5, 6, 14 ], "missing_information": "", "source_ids": [ "S01" ] }, { "dimension_id": "D1", "dimension_name": "过程透明度", "score": 60, "status": "rated", "confidence": "高", "judgment_range": null, "rationale": "划分超参脚本较清晰，α/β及8与16专家配置不一致。", "evidence_pages": [ 6, 7, 11, 12 ], "missing_information": "", "source_ids": [ "S01" ] }, { "dimension_id": "D2", "dimension_name": "研究资源可获得性", "score": 75, "status": "tentative", "confidence": "中", "judgment_range": [ 65, 85 ], "rationale": "官方readme入口、数据和脚本可访问，未执行完整复现。", "evidence_pages": [ 1 ], "missing_information": "见理由与材料边界", "source_ids": [ "S01" ] }, { "dimension_id": "D3", "dimension_name": "主张克制与报告一致性", "score": 40, "status": "rated", "confidence": "高", "judgment_range": null, "rationale": "持续最优、全MLP效率、插件同资源等表述偏宽，理论错误未限定。", "evidence_pages": [ 1, 5, 7, 10 ], "missing_information": "", "source_ids": [ "S01" ] }, { "dimension_id": "D4", "dimension_name": "表达清晰度", "score": 60, "status": "rated", "confidence": "高", "judgment_range": null, "rationale": "框图明确，但公式索引、符号和配置细节影响解读。", "evidence_pages": [ 3, 4, 10, 11, 12 ], "missing_information": "", "source_ids": [ "S01" ] }, { "dimension_id": "E1", "dimension_name": "复杂度合理性", "score": 65, "status": "rated", "confidence": "高", "judgment_range": null, "rationale": "三模块设计有动机，2048隐层稠密专家的必要性证据有限。", "evidence_pages": [ 4, 6, 12 ], "missing_information": "", "source_ids": [ "S01" ] }, { "dimension_id": "E2", "dimension_name": "资源效率与成本收益", "score": 60, "status": "tentative", "confidence": "中", "judgment_range": [ 45, 70 ], "rationale": "单场景优于重模型但慢于DLinear，缺完整训练与推理资源。", "evidence_pages": [ 7 ], "missing_information": "见理由与材料边界", "source_ids": [ "S01" ] }, { "dimension_id": "E3", "dimension_name": "可迁移与可复用程度", "score": 70, "status": "rated", "confidence": "高", "judgment_range": null, "rationale": "插件实验和入口利于复用，当前规格歧义增添成本。", "evidence_pages": [ 7, 12 ], "missing_information": "", "source_ids": [ "S01" ] }, { "dimension_id": "E4", "dimension_name": "后续研究启发性", "score": 75, "status": "rated", "confidence": "高", "judgment_range": null, "rationale": "促使区分尺度适应、容量和集成正则化，不把它转成用户任务。", "evidence_pages": [ 6, 14 ], "missing_information": "", "source_ids": [ "S01" ] }, { "dimension_id": "F1", "dimension_name": "与当前问题的相关性", "score": null, "status": "unknown", "confidence": "低", "judgment_range": null, "rationale": "未给出用户具体研究问题、已有知识、资源与当前决策，不能从论文的一般价值推断。", "evidence_pages": [], "missing_information": "个人背景未提供", "source_ids": [] }, { "dimension_id": "F2", "dimension_name": "相对于已有知识的学习收益", "score": null, "status": "unknown", "confidence": "低", "judgment_range": null, "rationale": "未给出用户具体研究问题、已有知识、资源与当前决策，不能从论文的一般价值推断。", "evidence_pages": [], "missing_information": "个人背景未提供", "source_ids": [] }, { "dimension_id": "F3", "dimension_name": "当前使用可行性", "score": null, "status": "unknown", "confidence": "低", "judgment_range": null, "rationale": "未给出用户具体研究问题、已有知识、资源与当前决策，不能从论文的一般价值推断。", "evidence_pages": [], "missing_information": "个人背景未提供", "source_ids": [] }, { "dimension_id": "F4", "dimension_name": "当前决策信息价值", "score": null, "status": "unknown", "confidence": "低", "judgment_range": null, "rationale": "未给出用户具体研究问题、已有知识、资源与当前决策，不能从论文的一般价值推断。", "evidence_pages": [], "missing_information": "个人背景未提供", "source_ids": [] } ], "risks": [ { "severity": "critical", "scope": "CL02", "issue": "式13→14漏d、21→22忽略列权重差，且后续周期索引条件缺失。", "pages": [ 10 ], "action": "暂停理论保证，不牵连独立实验事实。" }, { "severity": "major", "scope": "gate_specification", "issue": "表6 α=0代入式10逆向压低TopK，可能是α/β记号错误。", "pages": [ 4, 12 ], "action": "保留实现未知，不认定代码已反向训练。" }, { "severity": "major", "scope": "CL03", "issue": "ETTh2/720等明确不占优。", "pages": [ 5 ], "action": "收窄总体优越性。" } ], "aggregation": { "weights": null, "weighted_score_observed": null, "weighted_coverage": null, "tentative_weight_share": null, "notes": "保留24维原始分，不擅自合成总分；未知不作0。" }, "sources": [ { "source_id": "S01", "source_type": "arxiv PDF", "location": "/workspace/scratch/ac01e91af073/candidate_recovery/priority_packet_20261108/papers/ATT-0097/ATT-0097_paper.pdf", "version": "2406.03751v2" }, { "source_id": "S02", "source_type": "待核历史记录", "location": "/workspace/scratch/ac01e91af073/candidate_recovery/priority_packet_20261108/papers/ATT-0097/legacy_reference_only.json" } ], "limits": { "no_model_execution": true, "no_training_or_inference": true, "no_full_code_audit": true, "external_closest_prior_full_text_check": "本次未独立取得并阅读全文，原创性为暂定判断。", "author_result_not_independent_reproduction": true }, "resource_read": { "url": "https://github.com/TROUBADOUR000/AMD/blob/main/readme.md", "blob_sha": "0d0c44763ece014be578fffec439d26d6ad7b81c", "notes": "README.md大小写首次404，正确readme.md读取成功；仅资源入口未运行。" } }