Noise or Nuance: An Investigation Into Useful Information and Filtering For LLM Driven AKBC

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Clay, Alex, Jiménez-Ruiz, Ernesto, Madhyastha, Pranava
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866918139000782848
author Clay, Alex
Jiménez-Ruiz, Ernesto
Madhyastha, Pranava
author_facet Clay, Alex
Jiménez-Ruiz, Ernesto
Madhyastha, Pranava
contents RAG and fine-tuning are prevalent strategies for improving the quality of LLM outputs. However, in constrained situations, such as that of the 2025 LM-KBC challenge, such techniques are restricted. In this work we investigate three facets of the triple completion task: generation, quality assurance, and LLM response parsing. Our work finds that in this constrained setting: additional information improves generation quality, LLMs can be effective at filtering poor quality triples, and the tradeoff between flexibility and consistency with LLM response parsing is setting dependent.
format Preprint
id arxiv_https___arxiv_org_abs_2509_08903
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Noise or Nuance: An Investigation Into Useful Information and Filtering For LLM Driven AKBC
Clay, Alex
Jiménez-Ruiz, Ernesto
Madhyastha, Pranava
Computation and Language
RAG and fine-tuning are prevalent strategies for improving the quality of LLM outputs. However, in constrained situations, such as that of the 2025 LM-KBC challenge, such techniques are restricted. In this work we investigate three facets of the triple completion task: generation, quality assurance, and LLM response parsing. Our work finds that in this constrained setting: additional information improves generation quality, LLMs can be effective at filtering poor quality triples, and the tradeoff between flexibility and consistency with LLM response parsing is setting dependent.
title Noise or Nuance: An Investigation Into Useful Information and Filtering For LLM Driven AKBC
topic Computation and Language
url https://arxiv.org/abs/2509.08903