{ "rubric_version": "PAPER-RUBRIC-1.0", "reviewer": "E", "paper_id": "ATT-0367", "paper_title": "TimeDistill: Efficient Long-Term Time Series Forecasting with MLP via Cross-Architecture Distillation", "paper_version": "arXiv 2502.15016v3", "version_date": "2026/01/07", "first_public_date": "2025/02/20", "assessment_date": "2026-11-08", "reading_record": { "completed_at_utc": "2026-11-08T02:32:46.839645+00:00", "pdf_pages": 27, "coverage": "逐页全文及附录文本阅读；关键公式或表格按需原页视觉核对。参考文献已检查关联定位，未逐篇读取全部前作。", "experimental_execution": false }, "paper_type": [ "方法或算法", "实证分析" ], "analysis_mode": "standard", "user_context": { "research_question": null, "prior_knowledge": null, "constraints": null }, "legacy_review": { "read_before_new_review": true, "full_old_review_available": false, "history_counts_as_new_review": false, "record": { "candidate_id": "ATT-0367", "counts_as_review": false, "sidecar_files_supplied": false, "records": [] }, "handling": "旧完整评语及评分sidecar未提供；历史条目仅作为待核线索，不继承旧分数。" }, "claims": [ { "claim_id": "CL01", "claim": "跨架构多尺度多周期蒸馏提升轻量学生预测。", "claim_type": "性能", "conditions": "给定教师、八长预测基准、调优权重", "support_status": "部分支持，有负迁移例外", "source_ids": [ "S01" ] }, { "claim_id": "CL02", "claim": "训练后学生提供低参数低延迟部署。", "claim_type": "效率", "conditions": "不计教师训练，固定batch16与窗口720", "support_status": "部分支持，极值数字需校正", "source_ids": [ "S01" ] }, { "claim_id": "CL03", "claim": "周期与多尺度损失上界mixup增广损失并解释泛化。", "claim_type": "理论", "conditions": "定理4.1/4.2及附录A原式", "support_status": "存在直接反证；暂停作为可靠结论使用", "source_ids": [ "S01" ] } ], "scores": [ { "dimension_id": "A1", "dimension_name": "问题重要性", "score": 80, "status": "rated", "confidence": "高", "judgment_range": null, "rationale": "低延迟预测与训练端知识迁移是实际瓶颈。", "missing_information": "", "claim_ids": [], "source_ids": [ "S01" ], "location": "PDF p1—2" }, { "dimension_id": "A2", "dimension_name": "原创性", "score": 65, "status": "tentative", "confidence": "中", "judgment_range": [ 55, 75 ], "rationale": "多尺度周期跨架构组合有增量，广泛首创性未独立核对。", "missing_information": "多尺度周期跨架构组合有增量，广泛首创性未独立核对。", "claim_ids": [], "source_ids": [ "S01" ], "location": "PDF p2—5,18" }, { "dimension_id": "A3", "dimension_name": "实质推进程度", "score": 70, "status": "rated", "confidence": "高", "judgment_range": null, "rationale": "多数数据和教师下改善MLP，有负迁移和表格不一致限制。", "missing_information": "", "claim_ids": [], "source_ids": [ "S01" ], "location": "PDF p7,15,20" }, { "dimension_id": "A4", "dimension_name": "知识与解释增量", "score": 45, "status": "rated", "confidence": "高", "judgment_range": null, "rationale": "消融给出条件化线索，但mixup理论错误削弱机制认识。", "missing_information": "", "claim_ids": [], "source_ids": [ "S01" ], "location": "PDF p11—12,24—25" }, { "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": "MSE/MAE符合点预测目标，部署效率与训练成本区分不足。", "missing_information": "", "claim_ids": [], "source_ids": [ "S01" ], "location": "PDF p6—7,17" }, { "dimension_id": "B2", "dimension_name": "技术与论证正确性", "score": 35, "status": "rated", "confidence": "高", "judgment_range": null, "rationale": "训练目标可理解，定理A.2归一化错误可被零损失反例推翻。", "missing_information": "", "claim_ids": [], "source_ids": [ "S01" ], "location": "PDF p4—5,11—12" }, { "dimension_id": "B3", "dimension_name": "数据与评测流程完整性", "score": 70, "status": "tentative", "confidence": "中", "judgment_range": [ 60, 80 ], "rationale": "标准数据协议与drop_last统一，划分预处理多依赖外部实现。", "missing_information": "标准数据协议与drop_last统一，划分预处理多依赖外部实现。", "claim_ids": [], "source_ids": [ "S01" ], "location": "PDF p12—13" }, { "dimension_id": "B4", "dimension_name": "比较对象与比较公平性", "score": 75, "status": "rated", "confidence": "高", "judgment_range": null, "rationale": "强教师、简单学生、其他KD及归一化对照较齐，预算仍不等。", "missing_information": "", "claim_ids": [], "source_ids": [ "S01" ], "location": "PDF p7,13,18—19" }, { "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": "性能证据较多，但理论主张不成立且汇总消融存在冲突。", "missing_information": "", "claim_ids": [], "source_ids": [ "S01" ], "location": "PDF p7,11—12,24—25" }, { "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": "四组损失与真值替代消融有用，正则化和物理周期仍混杂。", "missing_information": "", "claim_ids": [], "source_ids": [ "S01" ], "location": "PDF p7,16,24—25" }, { "dimension_id": "C3", "dimension_name": "结果稳定性与不确定性处理", "score": 65, "status": "rated", "confidence": "高", "judgment_range": null, "rationale": "学生五种子误差小，基线匹配波动和逐horizon统计不全。", "missing_information": "", "claim_ids": [], "source_ids": [ "S01" ], "location": "PDF p14" }, { "dimension_id": "C4", "dimension_name": "适用范围与边界验证", "score": 70, "status": "rated", "confidence": "高", "judgment_range": null, "rationale": "八数据、多教师学生窗口，明确弱教师负迁移，缺少漂移边界。", "missing_information": "", "claim_ids": [], "source_ids": [ "S01" ], "location": "PDF p14—15,20,23" }, { "dimension_id": "D1", "dimension_name": "过程透明度", "score": 65, "status": "rated", "confidence": "高", "judgment_range": null, "rationale": "超参充分；理论尺寸、测速硬件及表格版本对应仍有缺口。", "missing_information": "", "claim_ids": [], "source_ids": [ "S01" ], "location": "PDF p11—14,17,24" }, { "dimension_id": "D2", "dimension_name": "研究资源可获得性", "score": 60, "status": "tentative", "confidence": "中", "judgment_range": [ 45, 80 ], "rationale": "公开代码入口明确，未核验核心文件、权重和可执行性。", "missing_information": "公开代码入口明确，未核验核心文件、权重和可执行性。", "claim_ids": [], "source_ids": [ "S01" ], "location": "PDF p1,12—13" }, { "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": "Win Keep、去监督优势、参数与加速比叙述有可核对冲突。", "missing_information": "", "claim_ids": [], "source_ids": [ "S01" ], "location": "PDF p2,7—8,17,24—25" }, { "dimension_id": "D4", "dimension_name": "表达清晰度", "score": 70, "status": "rated", "confidence": "高", "judgment_range": null, "rationale": "方法图和训练流程清楚，理论符号与附录数字增加阅读负担。", "missing_information": "", "claim_ids": [], "source_ids": [ "S01" ], "location": "PDF p4—5,11—12" }, { "dimension_id": "E1", "dimension_name": "复杂度合理性", "score": 70, "status": "rated", "confidence": "高", "judgment_range": null, "rationale": "增加训练损失而保留简单部署模型，周期项必要性依教师而变。", "missing_information": "", "claim_ids": [], "source_ids": [ "S01" ], "location": "PDF p5,24—25" }, { "dimension_id": "E2", "dimension_name": "资源效率与成本收益", "score": 65, "status": "tentative", "confidence": "中", "judgment_range": [ 50, 80 ], "rationale": "同batch延迟支持推理收益，训练成本缺失且倍数报告有误差。", "missing_information": "同batch延迟支持推理收益，训练成本缺失且倍数报告有误差。", "claim_ids": [], "source_ids": [ "S01" ], "location": "PDF p12,17,21" }, { "dimension_id": "E3", "dimension_name": "可迁移与可复用程度", "score": 80, "status": "rated", "confidence": "高", "judgment_range": null, "rationale": "教师冻结和多层监督接口可用于多种预测架构。", "missing_information": "", "claim_ids": [], "source_ids": [ "S01" ], "location": "PDF p4—6,20" }, { "dimension_id": "E4", "dimension_name": "后续研究启发性", "score": 70, "status": "rated", "confidence": "高", "judgment_range": null, "rationale": "教师误差互补与周期损失收益差异形成具体可检验问题。", "missing_information": "", "claim_ids": [], "source_ids": [ "S01" ], "location": "PDF p3,24—25" }, { "dimension_id": "F1", "dimension_name": "与当前问题的相关性", "score": null, "status": "unknown", "confidence": null, "judgment_range": null, "rationale": "未给定用户正在解决的具体预测或异常检测问题。", "missing_information": "个人研究背景缺失；不以论文质量代填。", "claim_ids": [], "source_ids": [], "location": "任务上下文（非论文证据）" }, { "dimension_id": "F2", "dimension_name": "相对于已有知识的学习收益", "score": null, "status": "unknown", "confidence": null, "judgment_range": null, "rationale": "未给定用户已掌握的理论、算法或已读论文。", "missing_information": "个人研究背景缺失；不以论文质量代填。", "claim_ids": [], "source_ids": [], "location": "任务上下文（非论文证据）" }, { "dimension_id": "F3", "dimension_name": "当前使用可行性", "score": null, "status": "unknown", "confidence": null, "judgment_range": null, "rationale": "未给定用户用途、算力、数据与许可约束。", "missing_information": "个人研究背景缺失；不以论文质量代填。", "claim_ids": [], "source_ids": [], "location": "任务上下文（非论文证据）" }, { "dimension_id": "F4", "dimension_name": "当前决策信息价值", "score": null, "status": "unknown", "confidence": null, "judgment_range": null, "rationale": "未给定当前待决策的具体方法或研究路线。", "missing_information": "个人研究背景缺失；不以论文质量代填。", "claim_ids": [], "source_ids": [], "location": "任务上下文（非论文证据）" } ], "risks": [ { "risk_id": "R01", "severity": "关键问题", "certainty": "已确认（原页核对及代数反例）", "description": "定理A.2将非归一化Q_y+ηQ_t当分布，log-sum后遗漏分母1+η。真值=教师=学生时左边0、右边(1+η)log(1+η)>0，原上界不成立；暂停CL03中该理论作为可靠结论使用。", "affected_claim_ids": [ "CL03" ], "source_ids": [ "S01" ], "location": "PDF p5,11—12" }, { "risk_id": "R02", "severity": "重要问题", "certainty": "已确认报告冲突；来源未知", "description": "主表与附录Traffic及部分消融均值不一致，Win Keep和去监督优越性叙述与表内反例冲突。", "affected_claim_ids": [ "CL01" ], "source_ids": [ "S01" ], "location": "PDF p7—8,24—25" }, { "risk_id": "R03", "severity": "重要问题", "certainty": "已确认报告冲突及证据缺口", "description": "参数数量跨图表不同，196倍加速与表11比值不符，教师训练与总预算未计入。", "affected_claim_ids": [ "CL02" ], "source_ids": [ "S01" ], "location": "PDF p2,12,17" } ], "aggregation": { "weights": null, "weighted_score_observed": null, "weighted_coverage": null, "tentative_weight_share": null, "numeric_dimensions": 20, "applicable_dimensions": 24, "unweighted_dimension_coverage": 0.8333333333333334, "notes": "保留24维原始分数，不计算论文总分或跨论文排序；F组未知。" }, "reading_recommendation": "定向查阅蒸馏实证；暂停将周期mixup上界作为可靠理论使用", "recommended_sections": [ "p4—5 蒸馏目标", "p24—25 双教师消融" ], "sources": [ { "source_id": "S01", "source_type": "paper_pdf", "title": "TimeDistill: Efficient Long-Term Time Series Forecasting with MLP via Cross-Architecture Distillation", "version_or_date": "2502.15016v3", "location": "/workspace/scratch/ac01e91af073/candidate_recovery/priority_packet_20261108/papers/ATT-0367/ATT-0367_paper.pdf", "sha256": "8f7b3e2db65e4a21253330bb8720204dfcbd4838ad500dde23e2a53408721d67", "access_status": "已读取完整PDF文本及附录；本地版本，不推定等同会议终稿" }, { "source_id": "S02", "source_type": "legacy_history", "location": "/workspace/scratch/ac01e91af073/candidate_recovery/priority_packet_20261108/papers/ATT-0367/legacy_reference_only.json", "access_status": "已先读；无旧完整评语" } ], "independence": "未读取同批其他评审的新意见。", "narrative_chinese_characters": 1730 }