Benchmarking and Improving Compositional Generalization of Multi-aspect Controllable Text Generation

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Zhong, Tianqi, Li, Zhaoyi, Wang, Quan, Song, Linqi, Wei, Ying, Lian, Defu, Mao, Zhendong
Format: Preprint
Veröffentlicht: 2024
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866909215215321088
author Zhong, Tianqi
Li, Zhaoyi
Wang, Quan
Song, Linqi
Wei, Ying
Lian, Defu
Mao, Zhendong
author_facet Zhong, Tianqi
Li, Zhaoyi
Wang, Quan
Song, Linqi
Wei, Ying
Lian, Defu
Mao, Zhendong
contents Compositional generalization, representing the model's ability to generate text with new attribute combinations obtained by recombining single attributes from the training data, is a crucial property for multi-aspect controllable text generation (MCTG) methods. Nonetheless, a comprehensive compositional generalization evaluation benchmark of MCTG is still lacking. We propose CompMCTG, a benchmark encompassing diverse multi-aspect labeled datasets and a crafted three-dimensional evaluation protocol, to holistically evaluate the compositional generalization of MCTG approaches. We observe that existing MCTG works generally confront a noticeable performance drop in compositional testing. To mitigate this issue, we introduce Meta-MCTG, a training framework incorporating meta-learning, where we enable models to learn how to generalize by simulating compositional generalization scenarios in the training phase. We demonstrate the effectiveness of Meta-MCTG through achieving obvious improvement (by at most 3.64%) for compositional testing performance in 94.4% cases.
format Preprint
id arxiv_https___arxiv_org_abs_2404_04232
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Benchmarking and Improving Compositional Generalization of Multi-aspect Controllable Text Generation
Zhong, Tianqi
Li, Zhaoyi
Wang, Quan
Song, Linqi
Wei, Ying
Lian, Defu
Mao, Zhendong
Computation and Language
Compositional generalization, representing the model's ability to generate text with new attribute combinations obtained by recombining single attributes from the training data, is a crucial property for multi-aspect controllable text generation (MCTG) methods. Nonetheless, a comprehensive compositional generalization evaluation benchmark of MCTG is still lacking. We propose CompMCTG, a benchmark encompassing diverse multi-aspect labeled datasets and a crafted three-dimensional evaluation protocol, to holistically evaluate the compositional generalization of MCTG approaches. We observe that existing MCTG works generally confront a noticeable performance drop in compositional testing. To mitigate this issue, we introduce Meta-MCTG, a training framework incorporating meta-learning, where we enable models to learn how to generalize by simulating compositional generalization scenarios in the training phase. We demonstrate the effectiveness of Meta-MCTG through achieving obvious improvement (by at most 3.64%) for compositional testing performance in 94.4% cases.
title Benchmarking and Improving Compositional Generalization of Multi-aspect Controllable Text Generation
topic Computation and Language
url https://arxiv.org/abs/2404.04232