OpenTSLM TeeMoE: A Unified Time-Series Language Model for Forecasting, Contextual Prediction, and Reasoning

2026-09-30

AI Summary

This study addresses the fragmentation of forecasting and contextual reasoning capabilities in existing time-series foundation models by proposing a universal time-series language model that seamlessly unifies these heterogeneous abilities. Methodologically, the approach employs a shared backbone network integrated with Low-Rank Adaptation (LoRA) and introduces a novel LoRA Mixture-of-Experts (MoE) controller. This controller dynamically fuses independently trained low-rank experts through weighted aggregation, enabling direct forecasting, text-based contextual reasoning, and the integration of external expert outputs. Experimental evaluations demonstrate that the proposed model achieves superior performance on mainstream benchmarks such as GIFT-Eval, ranking among the top three across multiple metrics. These results establish an effective paradigm for developing unified time-series agents.

Abstract

Real-world time-series applications increasingly require models that can handle time series forecasting, context-conditioned prediction, and language-based temporal reasoning. Yet current time-series foundation models remain fragmented across these capabilities: numerical specialists often provide the strongest forecasts, while language-based models offer broader contextual understanding and analysis. A central challenge is to unify these heterogeneous capabilities without reducing their individual performance. We introduce OpenTSLM TeeMoE, a generalist time-series language model that can forecast directly from observed time series, reason over textual context and temporal patterns, and synthesize and refine predictions from external numerical forecasting specialists. We independently train three low-rank experts for forecast aggregation, native forecasting, and temporal analysis over a shared backbone. A learned LoRA mixture-of-experts controller then weights their frozen parameter updates for each request. Our proposed model achieves strong performance on widely used benchmarks for time series forecasting, context-conditioned prediction, and language-based temporal reasoning, ranking among the top three on GIFT-Eval by mean MASE rank, Context is Key by RCRPS, and TimeSeriesExam by accuracy.

Problem

• Time-series forecasting
• Contextual prediction
• Temporal reasoning
• Foundation models
• Unified model

Innovation

• Time-Series Language Model
• Mixture-of-Experts
• LoRA
• Temporal Reasoning
• Forecast Aggregation