OpenTSLM TeeMoE: A Unified Time-Series Language Model for Forecasting, Contextual Prediction, and Reasoning
OpenTSLM TeeMoE is pitched as one model for forecasting, context-conditioned prediction, and temporal reasoning without giving up specialist performance.
The paper describes three independently trained low-rank experts: forecast aggregation, native forecasting, and temporal analysis, all sharing a backbone. A learned LoRA mixture-of-experts controller weights those frozen updates per request. The authors report top-three results on GIFT-Eval, Context is Key, and TimeSeriesExam under their cited metrics. ArXiv · AI/CL/LG's note
The paper describes three independently trained low-rank experts: forecast aggregation, native forecasting, and temporal analysis, all sharing a backbone. A learned LoRA mixture-of-experts controller weights those frozen updates per request. The authors report top-three results on GIFT-Eval, Context is Key, and TimeSeriesExam under their cited metrics. ArXiv · AI/CL/LG's note
score 5