Not All Retrievals are Useful: Cross-Attention for Input-Aware RAG in Time Series Forecasting
arXiv:2603.14709
The paper introduces Cross-RAG, a retrieval-augmented generation framework for zero-shot time series forecasting that uses query‑retrieval cross‑attention to selectively attend to relevant retrieved samples, improving accuracy and stability as the number of retrieved items grows.
Abstract
Retrieval-augmented generation (RAG) enhances zero-shot time series (TS) forecasting by leveraging external knowledge bases, yet existing approaches overlook input-level relevance when fusing retrieved samples with the query. We argue that not all retrievals are equally useful, and irrelevant ones can degrade performance. To this end, we propose Cross-RAG, a zero-shot RAG-based forecasting framework that selectively attends to query-relevant retrieved samples via query--retrieval cross-attention. By modeling input-level relevance between the query and retrieved samples, Cross-RAG jointly incorporates three sources of information: 1) the query itself, 2) the retrieved samples, and 3) their relational interactions. In particular, this input-aware design enables Cross-RAG to remain stable as the number of retrieved samples grows, whereas prior methods without cross-attention require careful tuning to avoid degradation from irrelevant retrievals. Extensive experiments demonstrate that Cross-RAG consistently improves zero-shot forecasting performance across multiple TSFM backbones and various RAG methods, with additional analyses confirming its effectiveness across various retrieval scenarios. Code is available at https://github.com/seunghan96/cross-rag/.
KDD Workshop on Mining and Learning from Time Series 2026 (Oral Presentation)