ConCISE: A Reference-Free Conciseness Evaluation Metric for LLM-Generated Answers

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Ghafari, Seyed Mohssen, Kol, Ronny, Quiroz, Juan C., Luan, Nella, Patial, Monika, Rupasinghe, Chanaka, Wandabwa, Herman, Pizzato, Luiz
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866914387317489664
author Ghafari, Seyed Mohssen
Kol, Ronny
Quiroz, Juan C.
Luan, Nella
Patial, Monika
Rupasinghe, Chanaka
Wandabwa, Herman
Pizzato, Luiz
author_facet Ghafari, Seyed Mohssen
Kol, Ronny
Quiroz, Juan C.
Luan, Nella
Patial, Monika
Rupasinghe, Chanaka
Wandabwa, Herman
Pizzato, Luiz
contents Large language models (LLMs) frequently generate responses that are lengthy and verbose, filled with redundant or unnecessary details. This diminishes clarity and user satisfaction, and it increases costs for model developers, especially with well-known proprietary models that charge based on the number of output tokens. In this paper, we introduce a novel reference-free metric for evaluating the conciseness of responses generated by LLMs. Our method quantifies non-essential content without relying on gold standard references and calculates the average of three calculations: i) a compression ratio between the original response and an LLM abstractive summary; ii) a compression ratio between the original response and an LLM extractive summary; and iii) wordremoval compression, where an LLM removes as many non-essential words as possible from the response while preserving its meaning, with the number of tokens removed indicating the conciseness score. Experimental results demonstrate that our proposed metric identifies redundancy in LLM outputs, offering a practical tool for automated evaluation of response brevity in conversational AI systems without the need for ground truth human annotations.
format Preprint
id arxiv_https___arxiv_org_abs_2511_16846
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle ConCISE: A Reference-Free Conciseness Evaluation Metric for LLM-Generated Answers
Ghafari, Seyed Mohssen
Kol, Ronny
Quiroz, Juan C.
Luan, Nella
Patial, Monika
Rupasinghe, Chanaka
Wandabwa, Herman
Pizzato, Luiz
Computation and Language
Artificial Intelligence
I.2.0
Large language models (LLMs) frequently generate responses that are lengthy and verbose, filled with redundant or unnecessary details. This diminishes clarity and user satisfaction, and it increases costs for model developers, especially with well-known proprietary models that charge based on the number of output tokens. In this paper, we introduce a novel reference-free metric for evaluating the conciseness of responses generated by LLMs. Our method quantifies non-essential content without relying on gold standard references and calculates the average of three calculations: i) a compression ratio between the original response and an LLM abstractive summary; ii) a compression ratio between the original response and an LLM extractive summary; and iii) wordremoval compression, where an LLM removes as many non-essential words as possible from the response while preserving its meaning, with the number of tokens removed indicating the conciseness score. Experimental results demonstrate that our proposed metric identifies redundancy in LLM outputs, offering a practical tool for automated evaluation of response brevity in conversational AI systems without the need for ground truth human annotations.
title ConCISE: A Reference-Free Conciseness Evaluation Metric for LLM-Generated Answers
topic Computation and Language
Artificial Intelligence
I.2.0
url https://arxiv.org/abs/2511.16846