Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

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

About

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.

Tianqi Zhong, Zhaoyi Li, Quan Wang, Linqi Song, Ying Wei, Defu Lian, Zhendong Mao• 2024

Related benchmarks

TaskDatasetResultRank
Multi-Aspect Controllable Text GenerationFyelp CompMCTG (Hold-Out)
Acomp94.06
12
Multi-Aspect Controllable Text GenerationFyelp ACD CompMCTG
Acomp90.4
12
Multi-Aspect Controllable Text GenerationCompMCTG 1.0 (Original)--
10
Multi-Aspect Controllable Text GenerationCompMCTG 1.0 (Hold-Out)--
10
Multi-Aspect Controllable Text GenerationCompMCTG 1.0 (ACD)--
10
Multi-Aspect Controllable Text GenerationCompMCTG Overall Summary Average 1.0--
10
Multi-attribute Conditional Text GenerationCompMCTG In-distribution Few-shot 1.0 (test)--
10
Multi-attribute Conditional Text GenerationCompMCTG Compositional Few-Shot 1.0 (test)--
10
Multi-attribute Controlled Text GenerationCompM-CTG Original--
10
Multi-attribute Controlled Text GenerationCompM-CTG (Hold-Out)--
10
Showing 10 of 15 rows

Other info

Code

Follow for update