Energy-Based Diffusion Language Models for Text Generation

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Xu, Minkai, Geffner, Tomas, Kreis, Karsten, Nie, Weili, Xu, Yilun, Leskovec, Jure, Ermon, Stefano, Vahdat, Arash
Format: Preprint
Veröffentlicht: 2024
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866909528349474816
author Xu, Minkai
Geffner, Tomas
Kreis, Karsten
Nie, Weili
Xu, Yilun
Leskovec, Jure
Ermon, Stefano
Vahdat, Arash
author_facet Xu, Minkai
Geffner, Tomas
Kreis, Karsten
Nie, Weili
Xu, Yilun
Leskovec, Jure
Ermon, Stefano
Vahdat, Arash
contents Despite remarkable progress in autoregressive language models, alternative generative paradigms beyond left-to-right generation are still being actively explored. Discrete diffusion models, with the capacity for parallel generation, have recently emerged as a promising alternative. Unfortunately, these models still underperform the autoregressive counterparts, with the performance gap increasing when reducing the number of sampling steps. Our analysis reveals that this degradation is a consequence of an imperfect approximation used by diffusion models. In this work, we propose Energy-based Diffusion Language Model (EDLM), an energy-based model operating at the full sequence level for each diffusion step, introduced to improve the underlying approximation used by diffusion models. More specifically, we introduce an EBM in a residual form, and show that its parameters can be obtained by leveraging a pretrained autoregressive model or by finetuning a bidirectional transformer via noise contrastive estimation. We also propose an efficient generation algorithm via parallel important sampling. Comprehensive experiments on language modeling benchmarks show that our model can consistently outperform state-of-the-art diffusion models by a significant margin, and approaches autoregressive models' perplexity. We further show that, without any generation performance drop, our framework offers a 1.3$\times$ sampling speedup over existing diffusion models. Reproduced code is available at https://github.com/MinkaiXu/Energy-Diffusion-LLM.
format Preprint
id arxiv_https___arxiv_org_abs_2410_21357
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Energy-Based Diffusion Language Models for Text Generation
Xu, Minkai
Geffner, Tomas
Kreis, Karsten
Nie, Weili
Xu, Yilun
Leskovec, Jure
Ermon, Stefano
Vahdat, Arash
Computation and Language
Machine Learning
Despite remarkable progress in autoregressive language models, alternative generative paradigms beyond left-to-right generation are still being actively explored. Discrete diffusion models, with the capacity for parallel generation, have recently emerged as a promising alternative. Unfortunately, these models still underperform the autoregressive counterparts, with the performance gap increasing when reducing the number of sampling steps. Our analysis reveals that this degradation is a consequence of an imperfect approximation used by diffusion models. In this work, we propose Energy-based Diffusion Language Model (EDLM), an energy-based model operating at the full sequence level for each diffusion step, introduced to improve the underlying approximation used by diffusion models. More specifically, we introduce an EBM in a residual form, and show that its parameters can be obtained by leveraging a pretrained autoregressive model or by finetuning a bidirectional transformer via noise contrastive estimation. We also propose an efficient generation algorithm via parallel important sampling. Comprehensive experiments on language modeling benchmarks show that our model can consistently outperform state-of-the-art diffusion models by a significant margin, and approaches autoregressive models' perplexity. We further show that, without any generation performance drop, our framework offers a 1.3$\times$ sampling speedup over existing diffusion models. Reproduced code is available at https://github.com/MinkaiXu/Energy-Diffusion-LLM.
title Energy-Based Diffusion Language Models for Text Generation
topic Computation and Language
Machine Learning
url https://arxiv.org/abs/2410.21357