ChronoScribe improves time series forecasting with dual AI specialization
ChronoScribe: Semantic-Numerical Specialization for Instruction-Driven Time Series Forecasting
Artificial IntelligenceMachine Learning
Summary
Forecasting future data points in time series is hard because it needs understanding natural language instructions and making exact numerical predictions. The authors show that using a two-part system, where a language AI interprets the task and a separate specialized model does the actual number crunching, works better than using just one AI for everything. They tested their approach called ChronoScribe on many datasets and found it predicted more accurately than other AI models. This method splits the job so each part focuses on what it does best.
What this means in practice
- •For business analysts: Generate precise forecasts from natural language queries without needing complex manual setup.
- •For software engineers: Integrate specialized forecasting systems that combine language understanding with accurate numerical models for flexible prediction tasks.
- •For financial modelers: Produce reliable forecast trajectories from human instructions, supporting investment and trading decision tools.$Commercial implications: Enables development of finance products offering accurate, instruction-driven forecasts for traders and portfolio managers.
Authors
Yuchen Jiao, Miao Zhang, Yuxuan Liu, Chengxi Li
Abstract
Recent advances in instruction-driven time series forecasting highlight a fundamental tension: systems must simultaneously master natural-language instruction following and precise numerical prediction. Current approaches often force a compromise. Large language models (LLMs) can interpret unstructured requests but may be inaccurate when asked to produce forecast values directly, whereas time series foundation models (TSFMs) provide numerical forecasts but typically require structured inputs specifying the target series, context window, and horizon. We present ChronoScribe, which applies semantic-numerical specialization to this workflow. A LoRA-adapted LLM proposes an execution path and context length; deterministic rules resolve the executable configuration. On the specialist path, a frozen TSFM produces quantile trajectories for the task-specified horizon. Deterministic trajectory summaries condition the response, while the LLM neither generates nor revises predicted values. We evaluate a Qwen3.5-9B--Chronos-2 implementation across 23 evaluation dimensions and 30 public datasets against eleven proprietary direct LLMs and three open TSFMs. ChronoScribe has the lowest error on sMAPE, MASE, WAPE, RMSE, and MAE among all evaluated systems, including fixed-context controls. Across these metrics, even the lowest direct-LLM error remains 2.0--2.7 times the corresponding ChronoScribe error. At the system level, ChronoScribe achieves a Task-Level Score (TLS) of 65.5, 10.6 points above the strongest direct-LLM baseline. The results support ChronoScribe's division of labor, in which the LLM handles interpretation, configuration, and communication while the TSFM handles numerical forecasting.