SACQ: Structured Decoding with Memory-Conditioned Refinement for Long-Horizon Forecasting
We present SACQ, a plug-in structured prediction head that replaces flatten readout while keeping the encoder unchanged.
Key points
- Long-term time series forecasting (LTSF) models predominantly employ patch-based encoders terminated by a flatten readout head that maps the entire encoded historical memory to all future steps through a single shared projection.
- This implicit coupling of future positions obscures position-specific historical-to-future alignment and amplifies sensitivity to corrupted inputs and extreme supervision noise.
- SACQ adopts a two-stage decoding pipeline: it first establishes a coarse patch-grid forecast scaffold, then refines each future position through cross-attention over historical memory and merges the attention-derived correction with the coarse scaffold via a learned per-patch gate.
- To stabilize optimization under long horizons and noisy labels, we further propose a batch-adaptive scaled log-cosh loss that automatically calibrates robustness to the current residual scale, suppressing outlier gradients while preserving MSE-like sensitivity for typical errors.
Sources (1)
- [1]SACQ: Structured Decoding with Memory-Conditioned Refinement for Long-Horizon ForecastingarXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 8, 03:25 AM
We present SACQ, a plug-in structured prediction head that replaces flatten readout while keeping the encoder unchanged.
Long-term time series forecasting (LTSF) models predominantly employ patch-based encoders terminated by a flatten readout head that maps the entire encoded historical memory to all future steps through a single shared projection.
Extractive summary: sentences quoted from the sources.
Before this
- Oct 7, 2026HAN-Mamba: Hierarchical Selective State Space Networks for Multi-Scale Financial Volatility Forecasting
- Oct 7, 2026FedSSMCoOp: SSM Encoders for light-weight Federated Prompt Learning for Few-shot Classification
- Oct 7, 2026GeoPrior-Mamba: Structured Process Priors with Mamba for Fine-Resolution XCO2 Reconstruction
- Oct 7, 2026EM-SNN: Efficiently Modulated Spiking Neural Network for Remote Sensing Image Dehazing
- Oct 6, 2026MaRK: Markov-adapted Recurrent Kernels for Dynamic Operator Conditioning in State Space Models
- Oct 6, 2026RSJEV: Discriminative Remote Sensing Scene Classification with Multimodal Large Language Models