time-series-forecasting

An arXiv paper presenting rc_prediction, an open-source Python package for parameter-aware reservoir computing (PARC) that learns parameter-dependent dynamics from data at a few operating points and predicts trajectories, bifurcations, and critical transitions at unseen parameter values.

论文提出 TSFLab，一个自我改进、与具体智能体解耦的时间序列 forecasting 研究工作台，为编码智能体提供共享资源与工具，以低成本完成复现、基准测试与新方法开发，并显著提升任务完成率、降低 token 与开发时间成本。

We introduce FACTS-MAS, a domain-agnostic multi-agent forecasting framework that decomposes heterogeneous evidence into specialized agent channels, validates and weights each via leakage-safe historical validation, and fuses them adaptively. Evaluated on Zillow housing inventory and FRED unemployment claims across 15 U.S. metros, it outperforms naive baselines and TimesFM, though a factor-adjusted exponential smoothing model matches its fused performance.

A UCL paper proposes MAPS, a model-agnostic metric for quantifying the intrinsic predictability of multivariable time series forecasting tasks, evaluated against 66 baseline measures on synthetic and real-world datasets with strong correlation to reference predictability (r=0.963).

TempShap proposes a cell-level, loss-based Shapley framework for time-series forecasting that produces attribution tensors revealing when a feature acts in the input window and how its influence varies across the forecast horizon, outperforming existing Shapley estimators on deletion-ranking benchmarks.

提出 Dual-Branch Calibration Module (DBCM)，通过共享适配器与低秩通道适配器的双分支设计，在时序预测的测试时自适应中处理非平稳分布偏移，并在多个数据集、预测跨度和骨干模型上超越现有方法。

该论文提出一种免训练的持续学习方法，通过GEPA式提示进化在冻结的LLM预测器（Llama-3.3-70B）的提示中演化约900词的引导文本，用于多模态时间序列预测，并在严格防数据污染的基准上超越TabPFN-TS-3等时间序列基础模型（GM MASE 0.776 vs 0.898）。

论文提出 ASE（Attribution-Guided Semantic Evolution）模块化框架，通过条件归因反馈引导语言模型构建与修订文本因子，并将其与数值预测器融合，在 TESS 数据集和 CSI 300 股票排序任务上取得改进。

该论文提出 Error Feedback Calibration（EFC），一种完全测试时的确定性后处理校正框架，通过 Causal Error Diagonal 与 Cumulative Error Transport 将已揭示的预测误差转化为对未来预测的因果校正，从而在多种时序预测骨干和数据集上提升性能并抑制过度校正。

TempoPFN 发表于 NeurIPS 2026，提出一个合成时间序列生成器与增强库，用于预训练线性 RNN 时间序列基础模型，实现零样本预测，并已开源代码与模型权重。

DST-Net 提出一种解耦的时空网络用于时间序列预测，针对趋势和季节分量分别设计潜在图块和高斯多尺度注意力机制，以匹配各分量的内在特性并在多个基准上验证有效。

该论文提出 Control-TS，通过残差控制将预测分位数作为分布控制信号引入预训练扩散预测模型，并提出 CLPS 指标评估生成未来轨迹的下游预测效用，在八个基准上一致提升 CRPS 表现。

该论文在一所大学建筑上对比了统计模型、XGBoost、LSTM与预训练时间序列基础模型TimesFM用于日前电力预测的效果，发现TimesFM 3.0配合协变量将平均绝对误差相对季节性朴素基线降低约43%，而XGBoost仍具竞争力。

本文提出 AliO（Align Outputs）方法，通过在时域和频域减少相同时间戳上预测输出的差异来提升长期时间序列预测模型的输出对齐性，并引入新指标 TAM（Time Alignment Metric）量化对齐程度，在保持或提升预测性能的同时最高提升 58.2% 的对齐度。

The paper presents HydroSphere, a governed data-driven framework for real-time wastewater management combining a TCN-LSTM forecasting model, PPO-based adaptive chemical dosing, and SHADE autoencoder anomaly detection, evaluated on 2.82 million water-quality measurements.

The paper introduces MACTS-EM, a multi-agent collaborative framework for time series forecasting that uses specialised agents, a meta-cognitive layer, emergent memory, and multimodal integration, reporting 8-12% accuracy gains and improved zero-shot transfer and regime-shift resilience across financial, climate, energy, and pandemic datasets.

This paper argues that time series forecasting benchmarks need scenario-grounded stress testing, where test cases include semantic scenarios, explicit failure operators, and measurable difficulty levels to capture structured anomalies like regime shifts and sensor failures that i.i.d. perturbation-based robustness tests miss.

A NeurIPS 2026 paper introduces topological out-of-domain generalization for dynamical systems reconstruction, identifying failure modes in hierarchical DSR models and fixing them with feature-splitting and physical sparsity priors, enabling correct prediction of bifurcations and regime changes without explicit control-parameter knowledge.

A new paper introduces CAGE, a framework combining generative models, adversarial discrimination, and conformal prediction to weight time series ensemble forecasts, showing improved robustness against outliers and noise on milk collection and monkeypox datasets.

The paper presents a comparative study of forecasting models for neutron monitor time series, including seasonal baselines and quantum-inspired architectures, highlighting that simple and functional models perform well on periodic scientific data.