Progress Ratio Embeddings: An Impatience Signal for Robust Length Control in Neural Text Generation

TL;DR

This paper introduces Progress Ratio Embeddings (PRE), a novel continuous length control method for neural text generation that enhances stability and generalization over existing approaches.

Contribution

The paper proposes PRE, a new length control technique using trigonometric signals, which improves robustness and generalization in Transformer-based text generation.

Findings

PRE provides stable length control without degrading accuracy.

PRE generalizes well to unseen target lengths.

Experiments on news-summarization benchmarks validate PRE's effectiveness.

Abstract

Modern neural language models achieve high accuracy in text generation, yet precise control over generation length remains underdeveloped. In this paper, we first investigate a recent length control method based on Reverse Positional Embeddings (RPE) and show its limits when control is requested beyond the training distribution. In particular, using a discrete countdown signal tied to the absolute remaining token count leads to instability. To provide robust length control, we introduce Progress Ratio Embeddings (PRE), as continuous embeddings tied to a trigonometric impatience signal. PRE integrates seamlessly into standard Transformer architectures, providing stable length fidelity without degrading text accuracy under standard evaluation metrics. We further show that PRE generalizes well to unseen target lengths. Experiments on two widely used news-summarization benchmarks validate these findings.