{ "rubric_version": "PAPER-RUBRIC-1.0", "paper_id": "ATT-0272", "paper_title": "MixLinear: Extreme Low Resource Multivariate Time Series Forecasting with $0.1K$ Parameters", "reviewer": "B", "analysis_mode": "standard", "paper_version": "2410.02081v2", "version_date": "2026/02/16", "first_public_date": "2024/10/02", "assessment_date": "2026-11-08", "reading_started_at": "2026-11-08T02:35:38.032213+00:00", "reading_completed_at": "2026-11-08T02:37:38.618823+00:00", "read_physical_pages": [ 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26 ], "visual_checks_physical_pages": [ 14, 21, 24, 25 ], "pdf_sha256": "bccc6a5a9d54f871fedb758c93424d09c02b6dfeb3dae3644e81bac3df622792", "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": "S5-272", "source_id": "S5", "source_library_file_id": "libfile_849759d66a808191813dbe9bd1605bb4", "source_filename": "已粘贴的文本(5).txt", "source_line": 295, "source_number": 272, "title": "MixLinear: Extreme Low Resource Multivariate Time Series Forecasting with $0.1K$ Parameters", "source_url": "https://proceedings.iclr.cc/paper_files/paper/2026/hash/7db2348b5bfeca620aa7327df815adcc-Abstract-Conference.html", "raw_line": "| P272 | [MixLinear: Extreme Low Resource Multivariate Time Series Forecasting with $0.1K$ Parameters](https://proceedings.iclr.cc/paper_files/paper/2026/hash/7db2348b5bfeca620aa7327df815adcc-Abstract-Conference.html) | 2 | 待处理 | 尚未评审；引用较低（2 次，低于 10 次门槛） |", "raw_cells": [ "P272", "[MixLinear: Extreme Low Resource Multivariate Time Series Forecasting with $0.1K$ Parameters](https://proceedings.iclr.cc/paper_files/paper/2026/hash/7db2348b5bfeca620aa7327df815adcc-Abstract-Conference.html)", "2", "待处理", "尚未评审；引用较低（2 次，低于 10 次门槛）" ], "normalized_title": "mixlinearextremelowresourcemultivariatetimeseriesforecastingwith01kparameters", "normalized_url": "proceedings.iclr.cc/paper_files/paper/2026/hash/7db2348b5bfeca620aa7327df815adcc-abstract-conference.html", "record_type": "legacy_unranked", "citation_verbatim": "2", "legacy_status_verbatim": "待处理", "legacy_unranked_reason_verbatim": "尚未评审；引用较低（2 次，低于 10 次门槛）", "counts_as_independent_review": false } ], "new_peer_reviews_read": false, "review_text": "## 独立判断\n本篇的低参数预测思路值得作为复现候选保留，但当前稿的公式、成本统计和实验叙述存在较多内部不一致，尚不能作为已验证的极低资源部署依据。先读历史记录，其中只有“未评审、引用较低”的清单行，不是完整旧评，也不据引用量扣质量分。本次阅读版本为二〇二六年二月的第二版，覆盖二十六页含附录，不能用最初预印本日期替代本版比较时间。方法将降采样后的序列分别送入时间分块线性变换和低秩频域变换，再相加恢复预测，且各通道共享参数。相较直接长度到长度的稠密线性层，这种结构限制确有减少参数的合理性；它结合了分块因子化与谱域瓶颈，属于实用结构组合，不宜将时频互补或低秩本身包装为全新理论。\n\n## 技术核查\n最需要先解决的是可实现的精确定义。正文第4页用长度n的频谱和n乘秩的复矩阵定义滤波，却给出以分段长度r计数的参数式，二者并非同一维度。附录第14页又把频域矩阵写成实数，并新增低通操作，未解释与正文的对应关系。该页式六声明两个投影矩阵均为平方根n乘d，却直接左乘平方根n乘平方根n的输入，通常维度不匹配；算法先把数据重排成n乘w，再重排成约n个元素的方阵，也缺少抽取、聚合和填充步骤。输出如何从方阵映射到任意预测长度、上采样是否含学习参数及卷积是否计入，均应由正式规格说明。这些问题不能自动证明仓库代码错误，但足以暂停对精确参数公式和复杂度推导的采信。\n\n线性层沿两个轴作用，并不意味着权重矩阵满足正交约束，全文也没有相应约束或证明。降采样本身可能滤掉高频，故时间分支捕捉局部尖锐变化的叙述需要更直接的证据。所谓低秩仅约束可学习线性算子的秩，不等同于证明原始时序的谱能量稀疏，低秩表现稳定也可能源自正则化或任务本身简单。第5页把二次降为近线性称为指数提升，在数学上不准确；总体计算还含FFT，必须与参数量的线性增长区分，并说明通道数、原始长度和预测长度的依赖。\n\n## 基线与核心结果\n主体采用八个常见长预测数据集、统一七百二十输入和四种预测长度，附录表五另含Weather但数据说明未列入。对SparseTSF、FITS等轻量基线的比较有价值：ETTh1的多个长度和Exchange中长预测确实更好，Traffic与Electricity则多项略差。ETTm1短预测、Weather以及Solar多项也不是最优，因此可接受的结论是参数约束下的竞争力，而非普遍最佳。表一的星号只是人为划定的改善百分比档，不是统计显著性。作者没有充分报告多种子方差或配对区间，小于百分之几的差异仍不稳定。正文说全部方法在同平台实施，附录又说基线结果取自FITS第一版；需要逐方法逐指标明确是引用值还是重新运行值，不能凭相同输入长度认定完全公平（第5—6、15、18页）。\n\n成本是核心卖点，却有直接可复核的问题。表一短预测九十六步中，MixLinear的MACs高于SparseTSF，例如ETTh1为167.66K对146.16K，否定“所有配置最低”的概括。图九把Electricity等标为K量级，而主表相近配置为M量级；部分其他图又反向混用。FITS的参数量正文给10512，附录表给41760，未说明配置变化。第7页Electricity延迟分别为2.05、4.20和4.77毫秒，相除约为2.05倍和2.33倍，并非文字的2.12倍及2.58倍。延迟缺批量、同步、预热和重复测量信息，A100八十GB上的结果也不能替代嵌入式设备内存、能耗和端到端延迟验证。\n\n## 消融、边界与使用建议\n去掉任一路径均有退化，尤其Traffic完整模型0.452对单时间路径0.528、单频率路径0.478，支持两者在当前设置互补；但移除分支也减少容量，没有等参数随机投影、纯时域低秩或相同降采样的单线性控制，尚未区分谱域表示与额外容量的贡献（第8页）。作者讨论高维数据的跨通道信息补偿，又明确模型为通道独立共享权重，只能谈跨通道训练样本共享，不能据此声称建模了通道间交互。分段长度图实际呈非单调成本曲线，正文却反复说随长度持续下降。预测可视化仅选择Exchange，附录第二图预测起点约六百而非所称七百二十，需要说明裁剪或索引；个案不能证明泛化。\n\n已核查官方README，存在环境、数据下载和训练脚本入口，未运行任何代码或测量参数量。建议先统一模型张量形状、精确实数自由参数计数及脚本版本，随后核验表格MAC单位和计时方法，再以等调参预算重跑SparseTSF与FITS并报告种子级结果；最终在目标低资源设备上验证真实收益。此文的研究启发是极小结构化预测器可能足够，但现稿尚未把“极少参数”“极低推理成本”“时频机制正确”三件事分别证明，评价应保留这个区分。", "review_hanzi_count": 1487, "claims": [ { "id": "CL01", "claim": "百量级参数下达到常见基准竞争精度", "judgment": "作者结果有条件支持，尚未独立核数", "pages": [ 6, 18 ] }, { "id": "CL02", "claim": "已严格说明参数与计算复杂度", "judgment": "暂停采信精确式，张量与统计口径需统一", "pages": [ 4, 5, 14, 21, 22 ] }, { "id": "CL03", "claim": "所有配置MAC最低且已适合边缘部署", "judgment": "前者有表内反例，后者缺目标设备验证", "pages": [ 6, 7, 22 ] } ], "scores": [ { "dimension_id": "A1", "dimension_name": "问题重要性", "score": 85, "status": "rated", "confidence": "高", "judgment_range": null, "rationale": "低资源长预测有明确部署需求。", "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": "时频双路径与因子化组合有增量，邻近方法全文未独立核读。", "evidence_pages": [ 2, 3, 9 ], "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": 60, "status": "rated", "confidence": "高", "judgment_range": null, "rationale": "约百参数运行点有意义，但成本记录与协议限制削弱推进证据。", "evidence_pages": [ 6, 18 ], "missing_information": "", "source_ids": [ "S01" ] }, { "dimension_id": "A4", "dimension_name": "知识与解释增量", "score": 45, "status": "rated", "confidence": "高", "judgment_range": null, "rationale": "稀疏、正交和跨通道解释超出消融支持。", "evidence_pages": [ 4, 8, 19, 21 ], "missing_information": "", "source_ids": [ "S01" ] }, { "dimension_id": "B1", "dimension_name": "问题定义与评价有效性", "score": 70, "status": "rated", "confidence": "高", "judgment_range": null, "rationale": "标准MSE任务清晰，部署目标却缺设备和能耗指标。", "evidence_pages": [ 2, 5, 7 ], "missing_information": "", "source_ids": [ "S01" ] }, { "dimension_id": "B2", "dimension_name": "技术与论证正确性", "score": 40, "status": "rated", "confidence": "高", "judgment_range": null, "rationale": "张量形状、参数公式与复杂度说法多处不一致；不等于已证代码错误。", "evidence_pages": [ 4, 5, 14 ], "missing_information": "", "source_ids": [ "S01" ] }, { "dimension_id": "B3", "dimension_name": "数据与评测流程完整性", "score": 55, "status": "rated", "confidence": "高", "judgment_range": null, "rationale": "有切分及归一化说明，Weather遗漏、基线来源及配置映射不清。", "evidence_pages": [ 15, 17, 18 ], "missing_information": "", "source_ids": [ "S01" ] }, { "dimension_id": "B4", "dimension_name": "比较对象与比较公平性", "score": 50, "status": "rated", "confidence": "高", "judgment_range": null, "rationale": "强轻量基线存在，但引用值与自跑值混合、等预算不明。", "evidence_pages": [ 5, 15, 18 ], "missing_information": "", "source_ids": [ "S01" ] }, { "dimension_id": "C1", "dimension_name": "核心主张的证据充分度", "score": 55, "status": "rated", "confidence": "高", "judgment_range": null, "rationale": "精度竞争力可见，普遍最优和精确成本优势不足。", "evidence_pages": [ 6, 7, 18, 22 ], "missing_information": "", "source_ids": [ "S01" ] }, { "dimension_id": "C2", "dimension_name": "机制归因与竞争解释区分度", "score": 50, "status": "rated", "confidence": "高", "judgment_range": null, "rationale": "双分支删除有效但缺容量匹配与时域低秩控制。", "evidence_pages": [ 8, 19, 21 ], "missing_information": "", "source_ids": [ "S01" ] }, { "dimension_id": "C3", "dimension_name": "结果稳定性与不确定性处理", "score": 35, "status": "rated", "confidence": "高", "judgment_range": null, "rationale": "没有充分多种子方差，星号仅为改善幅度分类。", "evidence_pages": [ 6, 18 ], "missing_information": "", "source_ids": [ "S01" ] }, { "dimension_id": "C4", "dimension_name": "适用范围与边界验证", "score": 50, "status": "rated", "confidence": "高", "judgment_range": null, "rationale": "多个数据和长度，但缺边缘实测、漂移缺失及跨领域迁移。", "evidence_pages": [ 5, 17, 24 ], "missing_information": "", "source_ids": [ "S01" ] }, { "dimension_id": "D1", "dimension_name": "过程透明度", "score": 45, "status": "rated", "confidence": "高", "judgment_range": null, "rationale": "伪代码难直接按形状实现，计时与复杂度规格欠缺。", "evidence_pages": [ 7, 14, 15 ], "missing_information": "", "source_ids": [ "S01" ] }, { "dimension_id": "D2", "dimension_name": "研究资源可获得性", "score": 70, "status": "tentative", "confidence": "中", "judgment_range": [ 55, 80 ], "rationale": "官方README确认环境数据及训练入口，未确认代码与论文逐项一致。", "evidence_pages": [ 11, 15 ], "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": 30, "status": "rated", "confidence": "高", "judgment_range": null, "rationale": "成本单位、速度算术与普遍领先叙述多处冲突。", "evidence_pages": [ 5, 6, 7, 18, 21, 22 ], "missing_information": "", "source_ids": [ "S01" ] }, { "dimension_id": "D4", "dimension_name": "表达清晰度", "score": 45, "status": "rated", "confidence": "高", "judgment_range": null, "rationale": "宏观架构直观，但术语夸张和正文附录不一致妨碍理解。", "evidence_pages": [ 3, 4, 14, 19 ], "missing_information": "", "source_ids": [ "S01" ] }, { "dimension_id": "E1", "dimension_name": "复杂度合理性", "score": 75, "status": "rated", "confidence": "高", "judgment_range": null, "rationale": "结构精简且适合强压缩，精确参数计数仍需厘清。", "evidence_pages": [ 3, 4, 14 ], "missing_information": "", "source_ids": [ "S01" ] }, { "dimension_id": "E2", "dimension_name": "资源效率与成本收益", "score": 45, "status": "tentative", "confidence": "中", "judgment_range": [ 30, 65 ], "rationale": "参数少有潜力，绝对MAC及边缘收益未核清。", "evidence_pages": [ 6, 7, 18, 21 ], "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": 60, "status": "rated", "confidence": "高", "judgment_range": null, "rationale": "共享通道结构易迁移，但输入预测尺寸规格及性能边界待查。", "evidence_pages": [ 14, 15 ], "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": "有价值的最小模型与等容量时频对照研究方向。", "evidence_pages": [ 8, 24 ], "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": "高", "issue": "公式维度和参数量不一致，精确复杂度主张待核", "pages": [ 4, 14 ] }, { "severity": "高", "issue": "MAC单位与速度算术不一致，成本主张待核", "pages": [ 6, 7, 18, 21, 22 ] }, { "severity": "中", "issue": "缺随机种子不确定性及统一基线来源", "pages": [ 15, 18 ] } ], "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-0272/ATT-0272_paper.pdf", "version": "2410.02081v2" }, { "source_id": "S02", "source_type": "待核历史记录", "location": "/workspace/scratch/ac01e91af073/candidate_recovery/priority_packet_20261108/papers/ATT-0272/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/aitianma/MixLinear/blob/main/README.md", "blob_sha": "8b231fca2649d4858cf9b290b518ca039e829e52", "execution_verified": false } }