SeaKR: Self-aware Knowledge Retrieval for Adaptive Retrieval Augmented Generation

Zijun Yao, Weijian Qi, Liangming Pan, Shulin Cao, Linmei Hu +3 more
6/27/2024
cs.CL

Official code: THU-KEG/SeaKR (43 stars, GPL-3.0, last commit 2024-06-26).

Abstract

This paper introduces Self-aware Knowledge Retrieval (SeaKR), a novel adaptive RAG model that extracts self-aware uncertainty of LLMs from their internal states. SeaKR activates retrieval when the LLMs present high self-aware uncertainty for generation. To effectively integrate retrieved knowledge snippets, SeaKR re-ranks them based on LLM's self-aware uncertainty to preserve the snippet that reduces their uncertainty to the utmost. To facilitate solving complex tasks that require multiple retrievals, SeaKR utilizes their self-aware uncertainty to choose among different reasoning strategies. Our experiments on both complex and simple Question Answering datasets show that SeaKR outperforms existing adaptive RAG methods. We release our code at https://github.com/THU-KEG/SeaKR.

View on arXivView PDF

Code Implementations(1)

THU-KEG/SeaKROfficial100%
GPL-3.0
437C, C++Jun 26, 20242 years agoGPL-3.0

Similar Papers

Adaptive Retrieval Without Self-Knowledge? Bringing Uncertainty Back Home

87% match1 repo212025

Self-Routing RAG: Binding Selective Retrieval with Knowledge Verbalization

86% match2025

SKILL-RAG: Self-Knowledge Induced Learning and Filtering for Retrieval-Augmented Generation

84% match2025

SUGAR: Leveraging Contextual Confidence for Smarter Retrieval

83% match2025

Knowing You Don't Know: Learning When to Continue Search in Multi-round RAG through Self-Practicing

82% match1 repo352025

Optimizing Knowledge Integration in Retrieval-Augmented Generation with Self-Selection

82% match2025

Cite this paper

@article{yao2024seakr,
  title  = {SeaKR: Self-aware Knowledge Retrieval for Adaptive Retrieval Augmented Generation},
  author = {Zijun Yao and Weijian Qi and Liangming Pan and Shulin Cao and Linmei Hu and Weichuan Liu and Lei Hou and Juanzi Li},
  year   = {2024},
  eprint = {2406.19215},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2406.19215v1}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2406.19215.svg)](https://codeofpaper.com/paper/2406.19215)

Discussion