EUVIMEDEUVIMED entdecken
Crossref

Modeling Uncertainty Trends for Timely Retrieval in Dynamic RAG

Bo Li, Tian Tian, Zhenghua Xu, Hao Cheng, Shikun Zhang, Wei Ye

Proceedings of the AAAI Conference on Artificial Intelligence · 14.03.2026 · 40 (37) · 31527-31535

DOI: 10.1609/aaai.v40i37.40418 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Proceedings of the AAAI Conference on Artificial Intelligence, online erschienen am 14.03.2026. Verfasst von Bo Li und 5 weiteren Autor:innen. Die Arbeit wurde bisher 3-mal zitiert (Crossref).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Dynamic retrieval-augmented generation (RAG) allows large language models (LLMs) to fetch external knowledge on demand, offering greater adaptability than static RAG. A central challenge in this setting lies in determining the optimal timing for retrieval. Existing methods often trigger retrieval based on low token-level confidence, which may lead to delayed intervention after errors have already propagated. We introduce Entropy-Trend Constraint (ETC), a training-free method that determines optimal retrieval timing by modeling the dynamics of token-level uncertainty. Specifically, ETC utilizes first- and second-order differences of the entropy sequence to detect emerging uncertainty trends, enabling earlier and more precise retrieval. Experiments on six QA benchmarks with three LLM backbones demonstrate that ETC consistently outperforms strong baselines while reducing retrieval frequency. ETC is particularly effective in domain-specific scenarios, exhibiting robust generalization capabilities. Ablation studies and qualitative analyses further confirm that trend-aware uncertainty modeling yields more effective retrieval timing. The method is plug-and-play, model-agnostic, and readily integrable into existing decoding pipelines. Implementation code is included in the supplementary materials.

Zitieren

APA

Li, B., Tian, T., Xu, Z., Cheng, H., Zhang, S., & Ye, W. (2026). Modeling Uncertainty Trends for Timely Retrieval in Dynamic RAG. Proceedings of the AAAI Conference on Artificial Intelligence, 40(37), 31527-31535. https://doi.org/10.1609/aaai.v40i37.40418

Vancouver

Li B, Tian T, Xu Z, Cheng H, Zhang S, Ye W. Modeling Uncertainty Trends for Timely Retrieval in Dynamic RAG. Proceedings of the AAAI Conference on Artificial Intelligence. 2026;40(37):31527-31535. doi:10.1609/aaai.v40i37.40418

BibTeX
@article{Li2026,
  author = {Li, Bo and Tian, Tian and Xu, Zhenghua and Cheng, Hao and Zhang, Shikun and Ye, Wei},
  title = {Modeling Uncertainty Trends for Timely Retrieval in Dynamic RAG},
  journal = {Proceedings of the AAAI Conference on Artificial Intelligence},
  year = {2026},
  volume = {40},
  number = {37},
  pages = {31527--31535},
  doi = {10.1609/aaai.v40i37.40418}
}
Kostenlos starten →