SEED: A Structural Encoder for Embedding-Driven Decoding in Time Series Prediction with LLMs

2025-06-25

AI Summary

Existing multivariate time series forecasting models struggle to jointly model structural dependencies among variables and perform semantic reasoning or task adaptation: structural encoders lack semantic generalization capability, while large language models (LLMs) cannot directly process raw time-series data. To address this, we propose SEED—a novel framework that enables the first synergistic modeling of structured encoding and frozen LLMs. Its core innovation is an embedding-driven decoding mechanism, which comprises chunk-aware encoding, projection alignment, and semantic reprogramming to map time-series segments into task-aware prototypes. This mechanism decouples representation learning from inference, bridging the gap between numerical time-series modeling and semantic understanding. Extensive experiments on multiple benchmark datasets demonstrate that SEED significantly outperforms strong baselines, exhibiting superior cross-task generalization and accurate capture of inter-variable structural dependencies.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: (Large) Language Models
• Planning, Routing, and Scheduling: Planning with Language Models

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information

Abstract

Multivariate time series forecasting requires models to simultaneously capture variable-wise structural dependencies and generalize across diverse tasks. While structural encoders are effective in modeling feature interactions, they lack the capacity to support semantic-level reasoning or task adaptation. Conversely, large language models (LLMs) possess strong generalization capabilities but remain incompatible with raw time series inputs. This gap limits the development of unified, transferable prediction systems. Therefore, we introduce SEED, a structural encoder for embedding-driven decoding, which integrates four stages: a token-aware encoder for patch extraction, a projection module that aligns patches with language model embeddings, a semantic reprogramming mechanism that maps patches to task-aware prototypes, and a frozen language model for prediction. This modular architecture decouples representation learning from inference, enabling efficient alignment between numerical patterns and semantic reasoning. Empirical results demonstrate that the proposed method achieves consistent improvements over strong baselines, and comparative studies on various datasets confirm SEED's role in addressing the structural-semantic modeling gap.

Problem

• Bridging structural-semantic gap in time series forecasting
• Enabling LLMs to process raw time series data
• Unifying feature interaction and task adaptation capabilities

Innovation

• Token-aware encoder extracts time series patches
• Projection aligns patches with LLM embeddings
• Semantic reprogramming maps patches to prototypes