LLaTiSA: Towards Difficulty-Stratified Time Series Reasoning from Visual Perception to Semantics

2026-04-19

AI Summary

This work addresses the challenges of ambiguous task formulation and the lack of standardized benchmarks in time series understanding for large language models. To this end, the authors propose a four-level cognitive complexity taxonomy for time series reasoning, construct HiTSR—a dataset comprising 83k samples—and introduce a novel multi-stage curriculum fine-tuning approach that integrates visual patterns with numerical tables, augmented with verification-oriented chain-of-thought annotations. The resulting model, LLaTiSA, demonstrates strong performance across diverse time series reasoning tasks, significantly enhancing the temporal awareness and cross-task generalization capabilities of vision-language models. Notably, LLaTiSA exhibits robust out-of-distribution generalization, underscoring its effectiveness in real-world scenarios where data distributions may shift.

Technology Category

• Machine Learning: Time-Series/Data Streams
• Natural Language Processing: (Large) Language Models
• Computer Vision: Large Vision Models

Application Category

• Search and Retrieval-Augmented AI: Large language models for search
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Comprehensive understanding of time series remains a significant challenge for Large Language Models (LLMs). Current research is hindered by fragmented task definitions and benchmarks with inherent ambiguities, precluding rigorous evaluation and the development of unified Time Series Reasoning Models(TSRMs). To bridge this gap, we formalize Time Series Reasoning (TSR) via a four-level taxonomy of increasing cognitive complexity. We introduce HiTSR, a hierarchical time series reasoning dataset comprising 83k samples with diverse task combinations and verified Chain-of-Thought (CoT) trajectories. Leveraging HiTSR, we propose LLaTiSA, a strong TSRM that integrates visualized patterns with precision-calibrated numerical tables to enhance the temporal perception of Vision-Language Models (VLMs). Through a multi-stage curriculum fine-tuning strategy, LLaTiSA achieves superior performance and exhibits robust out-of-distribution generalization across diverse TSR tasks and real-world scenarios. Our code is available at https://github.com/RainingNovember/LLaTiSA.

Problem

• Time Series Reasoning
• Large Language Models
• Benchmark Ambiguity
• Cognitive Complexity
• Temporal Perception

Innovation

• Time Series Reasoning
• Vision-Language Models
• Hierarchical Dataset
• Curriculum Fine-tuning
• Chain-of-Thought