In-sample Curriculum Learning by Sequence Completion for Natural Language Generation

TL;DR

This paper introduces an in-sample curriculum learning approach for natural language generation that progressively trains models from generating the last words to the entire sequence, improving performance across tasks.

Contribution

It proposes a task-agnostic in-sample curriculum learning method based on sequence completion, avoiding reliance on task-specific difficulty scoring.

Findings

Significant performance improvements over strong baselines

Effective generalization across multiple NLP tasks

Demonstrates the viability of in-sample curriculum learning

Abstract

Curriculum learning has shown promising improvements in multiple domains by training machine learning models from easy samples to hard ones. Previous works which either design rules or train models for scoring the difficulty highly rely on task-specific expertise, and cannot generalize. Inspired by the "easy-to-hard" intuition, we propose to do in-sample curriculum learning for natural language generation tasks. Our learning strategy starts training the model to generate the last few words, i.e., do sequence completion, and gradually extends to generate the whole output sequence. Comprehensive experiments show that it generalizes well to different tasks and achieves significant improvements over strong baselines.