InA-Probe: Instruction-Aware Active Probing for Time Series Forecasting with LLMs

2026-06-07

AI Summary

This work addresses the limitations of current large language models (LLMs) in time series forecasting, particularly their difficulty in capturing fine-grained, non-stationary patterns and adapting flexibly to task-specific intentions. The study introduces, for the first time, an active probing paradigm into LLM-based time series prediction, proposing an instruction-aware dynamic probing mechanism. By injecting multi-level instructions and generating adaptive queries, the method constructs sample-level probes that, combined with instruction-aware self-attention and temporal cross-attention, jointly guide the alignment of task intent with fine-grained semantic priors. Evaluated on seven real-world benchmarks, the approach outperforms state-of-the-art models and achieves up to a 37% reduction in prediction error under cross-domain zero-shot transfer, significantly enhancing generalization and reasoning capabilities in complex forecasting scenarios.

Abstract

Large Language Models (LLMs) have recently demonstrated impressive potential for time series forecasting. However, existing methods predominantly rely on passive modality alignment or static task reprogramming, which often fail to capture fine-grained, non-stationary temporal patterns or to adapt to nuanced task intents. In this paper, we propose Instruction-aware Active Probing (InA-Probe), which shifts the paradigm from passive alignment toward an active, instruction-driven probing mechanism. Specifically, we design a Multi-Level Instruction Injection mechanism that enriches the model with both global task objectives and fine-grained, patch-level semantic priors. Building on this, an Adaptive Query Generation module produces sample-specific probes that are dynamically modulated by the temporal context. These probes are then refined through a dual-stage attention process: they first internalize task-specific intents via Instruction-Aware Self-Attention, and subsequently interrogate the projected temporal representations through Temporal Cross-Attention to extract salient patterns. Comprehensive experiments on seven real-world benchmarks show that InA-Probe consistently outperforms state-of-the-art deep learning and LLM-based baselines, excelling in both one-for-all generalization and zero-shot transfer while reducing forecasting error by up to 37\% in challenging cross-domain scenarios. Ablation studies further confirm that the synergy between adaptive querying and fine-grained instructions is key to unlocking the reasoning power of LLMs for complex time series.

Problem

• time series forecasting
• Large Language Models
• non-stationary patterns
• task intent adaptation
• modality alignment

Innovation

• Instruction-Aware Active Probing
• Multi-Level Instruction Injection
• Adaptive Query Generation
• Temporal Cross-Attention
• LLM-based Time Series Forecasting