Skip to content

FLAME: Flow Enhanced Legendre Memory Models for General Time Series Forecasting

Dec 2025 · arXiv.org · Vol abs/2512.14253 · 4 citations · 115 references
Computer Science

Abstract

In this work, we introduce FLAME, a family of extremely lightweight and capable Time Series Foundation Models, which support versatile forecasting tasks via generative probabilistic modeling, while ensuring both efficiency and robustness. FLAME utilizes the Legendre Memory for strong generalization capabilities. By adapting variants of Legendre Memory, i.e., translated Legendre (LegT) and scaled Legendre (LegS), in the Encoding and Decoding phases, FLAME can effectively capture the inherent inductive bias within data and make efficient long-range inferences. To enhance the accuracy of probabilistic forecasting while remaining efficient, FLAME adopts a normalizing-flow-based forecasting head, which can model complex distributions over the forecasting horizon in a generative manner. Comprehensive experiments on three well-recognized benchmarks, including TSFM-Bench, ProbTS, and TFB, demonstrate that FLAME is a strong out-of-the-box tool for decision intelligence.

View source

Similar papers

Aurora-X: Built for Extreme Time Series Forecasting

A novel pattern-guided mixture-of-experts that expands model capacity through sparse activation and uses shallow patch similarities to constrain deep-layer routing, guiding expert specialization across heterogeneous time series and an implicit quantile network head that predicts arbitrary quantiles to characterize pred...

Xingjian Wu, Chen-Juan Guo, Xiangfei Qiu et al. · 0 citations
#reinforcement learning Open access Sep 2026

From Bellman to Real-Time: Extensions to Complex Weather Regimes, Physics-Informed Optimization, and Full-Scale Validation

In our earlier methodology paper, we introduced a hierarchical framework combining graph compression, Diffusion Convolutional Recurrent Neural Networks (DCRNNs), and Multi-Agent Reinforcement Learning (MARL) to approximate Bellman’s optimality principle for real-time energy system control, validated using Palm Springs,...

W. B. Lee, Anthony G. Constantinides · 0 citations
Preprint Sep 2026

FAST-ML: A Hybrid Physics-Machine Learning Framework for Tropical Cyclone Intensity Forecasting

FAST-ML demonstrates that observation-driven parameter learning within physically constrained dynamics simultaneously enhances accuracy, interpretability, and computational efficiency, and demonstrates that observation-driven parameter learning within physically constrained dynamics simultaneously enhances accuracy, in...

Shi-Jie Xiao, Jonathan Lin, Thomas Ehrmann et al. · 0 citations
#machine learning Preprint Oct 2026

F$^3$NO: Frequency-Decomposed Finite-Time Flow-map Neural Operators with Cross-Scale Conditioning

Neural operators enable fast PDE forecasting, but repeated predictions accumulate errors and fine-scale structures remain difficult to resolve. We introduce a frequency-decomposed finite-time flow-map neural operator (F$^3$NO) that leverages updated low-frequency features to guide nonlinear refinement of high-frequency...

Fan Wu, Cheng Jing, Kookjin Lee · 0 citations
#machine learning Preprint Oct 2026

Time-series Foundation Models for Predictive Control: The Role of Excitation

Deploying model predictive control (MPC) requires constructing or identifying a predictive model for each target system. Time-series foundation models (TSFMs) offer an attractive option thanks to strong zero-shot forecasting capabilities across systems. However, low forecast error does not guarantee that a TSFM capture...

Mazen Amria, Jasper Hoffmann, Philipp Bordne et al. · 0 citations
Open access Sep 2026

Machine Learning Post-Processing of Atmospheric River Persistence Forecasts: A Pre-Trained Tabular Transformer Across Mid-Latitude West Coasts

Atmospheric river (AR)-driven flooding is a natural hazard that causes severe damage in many regions, and the damage escalates sharply once an AR persists beyond a certain duration. Existing studies and numerical weather prediction models, however, have focused mainly on AR occurrence and on the intensity of integrated...

H. Chung, Cheong-Gi Kim · 0 citations

Related blog posts

Microsoft Research Blog Nov 16, 2026

TokenEconomizer: Learning from Experience to Make AI Coding Agents More Efficient

AI coding agents are increasingly help developers fix bugs and build features largely on their own, but every interaction with their models consumes tokens. In each session, an agent sends instructions, tool descriptions, and session history to its model. Most efforts to reduce this cost focus on individual tasks. Yet developers return to the same […] The post TokenEconomizer: Learning from Experience to Make AI Coding Agents More Efficient appeared first on Microsoft Research.

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.