{"id":6775,"date":"2026-05-02T03:30:10","date_gmt":"2026-05-02T03:30:10","guid":{"rendered":"https:\/\/scipapermill.com\/index.php\/2026\/05\/02\/time-series-forecasting-unpacking-the-latest-breakthroughs-from-mlps-to-llms-and-beyond\/"},"modified":"2026-05-02T03:30:10","modified_gmt":"2026-05-02T03:30:10","slug":"time-series-forecasting-unpacking-the-latest-breakthroughs-from-mlps-to-llms-and-beyond","status":"publish","type":"post","link":"https:\/\/scipapermill.com\/index.php\/2026\/05\/02\/time-series-forecasting-unpacking-the-latest-breakthroughs-from-mlps-to-llms-and-beyond\/","title":{"rendered":"Time Series Forecasting: Unpacking the Latest Breakthroughs \u2013 From MLPs to LLMs, and Beyond!"},"content":{"rendered":"

Latest 9 papers on time series forecasting: May. 2, 2026<\/h3>\n

Time series forecasting is the heartbeat of countless industries, from energy grids to financial markets, and fluid dynamics simulations. Accurately predicting future values from historical data is a challenge constantly pushing the boundaries of AI\/ML. Recent research reveals exciting advancements, tackling issues from model complexity and interpretability to real-world operational challenges and high-performance computing needs. Let\u2019s dive into some of the latest breakthroughs that are reshaping the landscape of time series prediction.<\/p>\n

The Big Idea(s) & Core Innovations<\/h3>\n

One striking theme emerging from recent work is the push for more efficient, robust, and interpretable models, sometimes even challenging the dominance of complex architectures. For instance, the paper \u201cITS-Mina: A Harris Hawks Optimization-Based All-MLP Framework with Iterative Refinement and External Attention for Multivariate Time Series Forecasting\u201d<\/a> by Pourya Zamanvaziri et al.\u00a0from Shahid Beheshti University, Iran<\/em>, introduces ITS-Mina<\/strong>, an all-MLP framework that surprisingly outperforms<\/em> Transformer-based models on several multivariate datasets. Their key insight? Simple MLP architectures, combined with iterative refinement via shared-parameter residual mixer loops and a linear-complexity external attention mechanism, can achieve state-of-the-art performance with less computational overhead. They even use Harris Hawks Optimization for automatic dropout tuning, showing that nature-inspired metaheuristics can optimize continuous hyperparameters effectively.<\/p>\n

Challenging the conventional wisdom in a different way, \u201cDecompKAN: Decomposed Patch-KAN for Long-Term Time Series Forecasting\u201d<\/a> by Naveen Mysore<\/em> introduces DecompKAN<\/strong>, an attention-free architecture leveraging Kolmogorov-Arnold Networks (KANs) for long-term forecasting. This model prioritizes interpretability<\/em>, allowing direct visualization of learned 1D nonlinear transformations via B-spline KAN edge functions. A core takeaway is that the overall pipeline design (decomposition, patching, normalization) often matters more than the specific nonlinear layer (KAN vs.\u00a0attention vs.\u00a0linear), especially for physics-structured data.<\/p>\n

However, the application of KANs to time series isn\u2019t without its caveats. \u201cAutocorrelation Reintroduces Spectral Bias in KANs for Time Series Forecasting\u201d<\/a> by Chen Zeng et al.\u00a0from Southeast University, China<\/em>, reveals a critical issue: KANs, despite theoretical claims of overcoming spectral bias under independent inputs, suffer from it in time series due to temporal autocorrelation. Their solution, DCT-KAN<\/strong>, uses Discrete Cosine Transform (DCT) preprocessing to decorrelate inputs, significantly reducing low-frequency preference and improving high-frequency component recovery.<\/p>\n

Meanwhile, Large Language Models (LLMs) are also stepping into the forecasting arena. \u201cCastFlow: Learning Role-Specialized Agentic Workflows for Time Series Forecasting\u201d<\/a> by Bokai Pan et al.\u00a0from the University of Science and Technology of China<\/em>, proposes CastFlow<\/strong>, an agentic framework that transforms forecasting from a static one-shot generation to a dynamic, multi-step workflow. It uses role-specialized LLMs (a frozen LLM for reasoning, a fine-tuned one for numerical forecasting) with an evidence-guided iterative refinement process, demonstrating how LLMs can bring advanced reasoning to numerical tasks. The framework is trained using a two-stage workflow combining supervised fine-tuning (SFT) and reinforcement learning with verifiable rewards (RLVR).<\/p>\n

Extending on the success of Mamba models, \u201cAdaMamba: Adaptive Frequency-Gated Mamba for Long-Term Time Series Forecasting\u201d<\/a> by Xudong Jiang et al.\u00a0from Tongji University<\/em>, addresses cross-domain heterogeneity. AdaMamba<\/strong> endogenizes adaptive frequency-domain analysis directly into Mamba\u2019s state-space update. By extending the temporal forgetting gate to a unified time-frequency forgetting gate, it dynamically calibrates state transitions based on learned frequency importance, while retaining Mamba\u2019s linear complexity.<\/p>\n

Under the Hood: Models, Datasets, & Benchmarks<\/h3>\n

This collection of papers highlights several crucial models, datasets, and benchmarking initiatives:<\/p>\n