ESQA: Event Sequences Question Answering

IEEE Access. 2026. Vol. 14. P. 59390–59408.

Abdullaeva I., Karpukhin I., Filatov A., Orlov M., Vasilev V., Belova N. S., Dimitrov D., Kuznetsov A., Kireev I., Savchenko A.

Event sequences, a specialized type of tabular data annotated with timestamps, are prevalent across practical domains such as finance, retail, social networks, and healthcare. Despite the importance of event sequence modeling and analysis, there has been little effort to adapt Large Language Models (LLMs) to this domain. In this paper, we propose a novel solution for event sequence processing that can address multiple downstream tasks with minimal or no fine-tuning. Specifically, we tackle the challenges of handling long sequences and enhance the processing of temporal and numeric features. Furthermore, we employ question-answering techniques to unify all downstream tasks within a single framework. The resulting method, called ESQA, effectively leverages the power of LLMs and, according to extensive experiments, achieves state-of-the-art performance across multiple fintech applications.

Keywords: large language model (LLM), event sequences, большие языковые модели (LLM), tabular temporal data, zero-shot recognition, question-answering, последовательность событий, табличные временные данные