EUVIMEDEUVIMED entdecken
Crossref

BRIEF: Bi-Level Coreset Selection for Efficient Instruction Tuning in LLMs

Chaoyuan Shen, Chi Zhang, Chengliang Chai, Jiacheng Wang, Jia Yuan, Yuping Wang, Ye Yuan, Guoren Wang, Lei Cao

Proceedings of the VLDB Endowment · 01.02.2026 · 19 (6) · 1264-1277

DOI: 10.14778/3797919.3797933 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Proceedings of the VLDB Endowment, online erschienen am 08.05.2026. Verfasst von Chaoyuan Shen und 8 weiteren Autor:innen. Die Arbeit nennt 73 Literaturangaben.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Instruction tuning is a key step in adapting large language models (LLMs) to effectively understand and follow human instructions. It enables LLMs to transform general knowledge into task-specific responses that align with user intent. Although many high-quality instruction tuning datasets have been released, efficiently utilizing these data sources during supervised fine-tuning (SFT) is important, as training on the full high-quality corpus can be computationally expensive. To address this inefficiency, we explore whether a compact, high-quality subset of instruction data can achieve comparable performance to full-dataset SFT, thereby reducing training cost without sacrificing effectiveness. To this end, this work proposes to select such a subset (a.k.a., coreset) of instruction examples that maintains comparable downstream performance while improving training efficiency. The key idea is inspired by our discovered decomposition that in instruction tuning, the training loss can be decomposed into two components that effectively quantify the contribution of an instruction to the two fundamental capabilities of LLMs, namely knowledge-related capability and instruction following capability. We then revisit the objective of the classical coreset approaches to balance the two capabilities when selecting instruction examples. Based on a bi-level formulation and a composite gradient distance that makes the objective submodular, we design an effective algorithm to achieve a bounded approximation error. Experiments on 4 datasets across 9 downstream tasks demonstrate that BRIEF reduces computational costs by 3× while improving accuracy by 5% on Llama-3.1-8B, Qwen3-4B and Mistral-Nemo-12B.

Zitieren

APA

Shen, C., Zhang, C., Chai, C., Wang, J., Yuan, J., Wang, Y., Yuan, Y., Wang, G., & Cao, L. (2026). BRIEF: Bi-Level Coreset Selection for Efficient Instruction Tuning in LLMs. Proceedings of the VLDB Endowment, 19(6), 1264-1277. https://doi.org/10.14778/3797919.3797933

Vancouver

Shen C, Zhang C, Chai C, Wang J, Yuan J, Wang Y, et al. BRIEF: Bi-Level Coreset Selection for Efficient Instruction Tuning in LLMs. Proceedings of the VLDB Endowment. 2026;19(6):1264-1277. doi:10.14778/3797919.3797933

BibTeX
@article{Shen2026,
  author = {Shen, Chaoyuan and Zhang, Chi and Chai, Chengliang and Wang, Jiacheng and Yuan, Jia and Wang, Yuping and Yuan, Ye and Wang, Guoren and Cao, Lei},
  title = {BRIEF: Bi-Level Coreset Selection for Efficient Instruction Tuning in LLMs},
  journal = {Proceedings of the VLDB Endowment},
  year = {2026},
  volume = {19},
  number = {6},
  pages = {1264--1277},
  doi = {10.14778/3797919.3797933}
}
Kostenlos starten →