EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Agentic Knowledge Curation Versus Full-Context Retrieval: An Empirical Study of Retrieval Failure Topology in Long-Context LLM Systems

Carlos A. Martín, Jesús M. Torres, Rosa M. Aguilar, Silvia Alayón, Manuel A. Bacallado

Applied Sciences · 06.07.2026 · 16 (13) · 6793

DOI: 10.3390/app16136793 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Applied Sciences, online erschienen am 06.07.2026. Verfasst von Carlos A. Martín und 4 weiteren Autor:innen. Die Arbeit nennt 42 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: CajaCanarias Foundation, La Caixa Foundation.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Karpathy’s proposal to replace Retrieval-Augmented Generation (RAG) with plain-text knowledge bases maintained directly by language agents (Agentic Knowledge Curation) has gained traction in industrial applications yet lacks systematic empirical evaluation. To our knowledge, this study presents the first comparative evaluation of this paradigm using blind human assessment by two independent external reviewers. The corpus comprises the technical documentation of a production semantic search system for Spanish legal documents (11 files, ~19,300 tokens) alongside 100 questions verified against source code, distributed across direct retrieval, multi-document synthesis, and reasoning about absence. Claude 3.5 Sonnet was utilized as the reference model to isolate the retrieval architecture’s effect. While overall accuracy was statistically indistinguishable between paradigms, agentic curation showed a significant advantage in direct retrieval over long documents. Conversely, error analysis revealed RAG false negatives associated with the lost-in-the-middle phenomenon, whereas agentic curation exhibited localized degradation in textual fidelity for queries requiring exact reproduction of formulas or sequences. These results characterize the differential error profiles of both paradigms, providing actionable design criteria for engineers managing technical documentation with language models in cloud environments.

Förderung

  • CajaCanarias Foundation
  • La Caixa Foundation

Zitieren

APA

Martín, C. A., Torres, J. M., Aguilar, R. M., Alayón, S., & Bacallado, M. A. (2026). Agentic Knowledge Curation Versus Full-Context Retrieval: An Empirical Study of Retrieval Failure Topology in Long-Context LLM Systems. Applied Sciences, 16(13), 6793. https://doi.org/10.3390/app16136793

Vancouver

Martín CA, Torres JM, Aguilar RM, Alayón S, Bacallado MA. Agentic Knowledge Curation Versus Full-Context Retrieval: An Empirical Study of Retrieval Failure Topology in Long-Context LLM Systems. Applied Sciences. 2026;16(13):6793. doi:10.3390/app16136793

BibTeX
@article{Martn2026,
  author = {Martín, Carlos A. and Torres, Jesús M. and Aguilar, Rosa M. and Alayón, Silvia and Bacallado, Manuel A.},
  title = {Agentic Knowledge Curation Versus Full-Context Retrieval: An Empirical Study of Retrieval Failure Topology in Long-Context LLM Systems},
  journal = {Applied Sciences},
  year = {2026},
  volume = {16},
  number = {13},
  pages = {6793},
  doi = {10.3390/app16136793}
}
Kostenlos starten →