Conditional [MASK] Discrete Diffusion Language Model

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Koh, Hyukhun, Jhang, Minha, Kim, Dohyung, Lee, Sangmook, Jung, Kyomin
Natura: Preprint
Pubblicazione: 2024
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866912243366494208
author Koh, Hyukhun
Jhang, Minha
Kim, Dohyung
Lee, Sangmook
Jung, Kyomin
author_facet Koh, Hyukhun
Jhang, Minha
Kim, Dohyung
Lee, Sangmook
Jung, Kyomin
contents Although auto-regressive models excel in natural language processing, they often struggle to generate diverse text and provide limited controllability. Non-auto-regressive methods could be an alternative but often produce degenerate outputs and exhibit shortcomings in conditional generation. To address these challenges, we propose Diffusion-EAGS, a novel framework that integrates conditional masked language models into diffusion language models through the theoretical lens of a conditional Markov Random Field. In doing so, we propose entropy-adaptive Gibbs sampling and entropy-based noise scheduling to counterbalance each model's shortcomings. Experimental results show that Diffusion-EAGS outperforms baselines and achieves the best quality-diversity tradeoff, demonstrating its effectiveness in non-autoregressive text generation.
format Preprint
id arxiv_https___arxiv_org_abs_2411_06438
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Conditional [MASK] Discrete Diffusion Language Model
Koh, Hyukhun
Jhang, Minha
Kim, Dohyung
Lee, Sangmook
Jung, Kyomin
Computation and Language
Artificial Intelligence
Although auto-regressive models excel in natural language processing, they often struggle to generate diverse text and provide limited controllability. Non-auto-regressive methods could be an alternative but often produce degenerate outputs and exhibit shortcomings in conditional generation. To address these challenges, we propose Diffusion-EAGS, a novel framework that integrates conditional masked language models into diffusion language models through the theoretical lens of a conditional Markov Random Field. In doing so, we propose entropy-adaptive Gibbs sampling and entropy-based noise scheduling to counterbalance each model's shortcomings. Experimental results show that Diffusion-EAGS outperforms baselines and achieves the best quality-diversity tradeoff, demonstrating its effectiveness in non-autoregressive text generation.
title Conditional [MASK] Discrete Diffusion Language Model
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2411.06438