Reranking-based Generation for Unbiased Perspective Summarization

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Ri, Narutatsu, Deas, Nicholas, McKeown, Kathleen
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866912439789944832
author Ri, Narutatsu
Deas, Nicholas
McKeown, Kathleen
author_facet Ri, Narutatsu
Deas, Nicholas
McKeown, Kathleen
contents Generating unbiased summaries in real-world settings such as political perspective summarization remains a crucial application of Large Language Models (LLMs). Yet, existing evaluation frameworks rely on traditional metrics for measuring key attributes such as coverage and faithfulness without verifying their applicability, and efforts to develop improved summarizers are still nascent. We address these gaps by (1) identifying reliable metrics for measuring perspective summary quality, and (2) investigating the efficacy of LLM-based methods beyond zero-shot inference. Namely, we build a test set for benchmarking metric reliability using human annotations and show that traditional metrics underperform compared to language model-based metrics, which prove to be strong evaluators. Using these metrics, we show that reranking-based methods yield strong results, and preference tuning with synthetically generated and reranking-labeled data further boosts performance. Our findings aim to contribute to the reliable evaluation and development of perspective summarization methods.
format Preprint
id arxiv_https___arxiv_org_abs_2506_15925
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Reranking-based Generation for Unbiased Perspective Summarization
Ri, Narutatsu
Deas, Nicholas
McKeown, Kathleen
Computation and Language
Generating unbiased summaries in real-world settings such as political perspective summarization remains a crucial application of Large Language Models (LLMs). Yet, existing evaluation frameworks rely on traditional metrics for measuring key attributes such as coverage and faithfulness without verifying their applicability, and efforts to develop improved summarizers are still nascent. We address these gaps by (1) identifying reliable metrics for measuring perspective summary quality, and (2) investigating the efficacy of LLM-based methods beyond zero-shot inference. Namely, we build a test set for benchmarking metric reliability using human annotations and show that traditional metrics underperform compared to language model-based metrics, which prove to be strong evaluators. Using these metrics, we show that reranking-based methods yield strong results, and preference tuning with synthetically generated and reranking-labeled data further boosts performance. Our findings aim to contribute to the reliable evaluation and development of perspective summarization methods.
title Reranking-based Generation for Unbiased Perspective Summarization
topic Computation and Language
url https://arxiv.org/abs/2506.15925