SOL: Measuring Gaps between Text Distributions by Double Sliced Wasserstein Metrics

Abstract

Evaluating text generation requires measuring how well the generated distribution matches the data distribution. For autoregressive models, this is done by the perplexity. Diffusion and flow-based language models can only provide a likelihood bound, whose tightness differs between model families. Sample-based substitutes such as generative perplexity with entropy do not consider the distribution fit. We propose SOL, a distance between text distributions. Each sequence is represented by the empirical measure of its hidden states under a fixed transformer and the distributions of these measures are compared by the double sliced Wasserstein distance. We prove that SOL is a metric if the transformer is injective. Experiments show that SOL detects distributional failures, recovers expected model trends, and provides stable sample-based estimates. We put forward SOL to fill the gap in the current evaluation protocol used for non auto-regressive models. As a first step we use SOL to re-evaluate a variety of models trained on OpenWebText.

The authors’ abstract, as arXiv lists it. arXiv’s listings, abstracts included, are free for anyone to reuse.

More in Computation & Language / NLP

• arXiv preprint · 5 October 2026

  Two-Sample Testing via Path-based Inference

  Eshant English, Wei-Cheng Lai, Yanfeng Yang and 3 others
• arXiv preprint · 6 October 2026

  Is $\sqrt{d}$ Separation Necessary for Gradient EM to Learn Gaussian Mixtures in High Dimensions?

  Yiran Zhang, Mo Zhou, Weihang Xu and 2 others
• arXiv preprint · 5 October 2026

  From Pixels, Without Pre-training: Joint Generative and Self-Supervised Representation Learning in One Model

  Vicente Balmaseda, Ching-Long Lin and Tianbao Yang
• arXiv preprint · 6 October 2026

  Learning a Mixture of GFlowNets

  Tiago da Silva, Amauri H. Souza and Salem Lahlou
• arXiv preprint · 5 October 2026

  Retrieval-Based In-Context Learning: A Domain Adaptation Framework

  Yilun Zhu, Naihao Deng, Yingcong Li and 2 others

All new papers in Computation & Language / NLP