Open-Domain Text Evaluation via Contrastive Distribution Methods

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Lu, Sidi, Liu, Hongyi, Celikyilmaz, Asli, Wang, Tianlu, Peng, Nanyun
Format: Preprint
Veröffentlicht: 2023
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866916280520409088
author Lu, Sidi
Liu, Hongyi
Celikyilmaz, Asli
Wang, Tianlu
Peng, Nanyun
author_facet Lu, Sidi
Liu, Hongyi
Celikyilmaz, Asli
Wang, Tianlu
Peng, Nanyun
contents Recent advancements in open-domain text generation, driven by the power of large pre-trained language models (LLMs), have demonstrated remarkable performance. However, assessing these models' generation quality remains a challenge. In this paper, we introduce a novel method for evaluating open-domain text generation called Contrastive Distribution Methods (CDM). Leveraging the connection between increasing model parameters and enhanced LLM performance, CDM creates a mapping from the _contrast_ of two probabilistic distributions -- one known to be superior to the other -- to quality measures. We investigate CDM for open-domain text generation evaluation under two paradigms: 1) _Generative_ CDM, which harnesses the contrast of two language models' distributions to generate synthetic examples for training discriminator-based metrics; 2) _Discriminative_ CDM, which directly uses distribution disparities between two language models for evaluation. Our experiments on coherence evaluation for multi-turn dialogue and commonsense evaluation for controllable generation demonstrate CDM's superior correlate with human judgment than existing automatic evaluation metrics, highlighting the strong performance and generalizability of our approach.
format Preprint
id arxiv_https___arxiv_org_abs_2306_11879
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle Open-Domain Text Evaluation via Contrastive Distribution Methods
Lu, Sidi
Liu, Hongyi
Celikyilmaz, Asli
Wang, Tianlu
Peng, Nanyun
Computation and Language
Recent advancements in open-domain text generation, driven by the power of large pre-trained language models (LLMs), have demonstrated remarkable performance. However, assessing these models' generation quality remains a challenge. In this paper, we introduce a novel method for evaluating open-domain text generation called Contrastive Distribution Methods (CDM). Leveraging the connection between increasing model parameters and enhanced LLM performance, CDM creates a mapping from the _contrast_ of two probabilistic distributions -- one known to be superior to the other -- to quality measures. We investigate CDM for open-domain text generation evaluation under two paradigms: 1) _Generative_ CDM, which harnesses the contrast of two language models' distributions to generate synthetic examples for training discriminator-based metrics; 2) _Discriminative_ CDM, which directly uses distribution disparities between two language models for evaluation. Our experiments on coherence evaluation for multi-turn dialogue and commonsense evaluation for controllable generation demonstrate CDM's superior correlate with human judgment than existing automatic evaluation metrics, highlighting the strong performance and generalizability of our approach.
title Open-Domain Text Evaluation via Contrastive Distribution Methods
topic Computation and Language
url https://arxiv.org/abs/2306.11879