Your LLM Knows the Future: Uncovering Its Multi-Token Prediction Potential

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Samragh, Mohammad, Kundu, Arnav, Harrison, David, Nishu, Kumari, Naik, Devang, Cho, Minsik, Farajtabar, Mehrdad
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866916845348454400
author Samragh, Mohammad
Kundu, Arnav
Harrison, David
Nishu, Kumari
Naik, Devang
Cho, Minsik
Farajtabar, Mehrdad
author_facet Samragh, Mohammad
Kundu, Arnav
Harrison, David
Nishu, Kumari
Naik, Devang
Cho, Minsik
Farajtabar, Mehrdad
contents Autoregressive language models are constrained by their inherently sequential nature, generating one token at a time. This paradigm limits inference speed and parallelism, especially during later stages of generation when the direction and semantics of text are relatively certain. In this work, we propose a novel framework that leverages the inherent knowledge of vanilla autoregressive language models about future tokens, combining techniques to realize this potential and enable simultaneous prediction of multiple subsequent tokens. Our approach introduces several key innovations: (1) a masked-input formulation where multiple future tokens are jointly predicted from a common prefix; (2) a gated LoRA formulation that preserves the original LLM's functionality, while equipping it for multi-token prediction; (3) a lightweight, learnable sampler module that generates coherent sequences from the predicted future tokens; (4) a set of auxiliary training losses, including a consistency loss, to enhance the coherence and accuracy of jointly generated tokens; and (5) a speculative generation strategy that expands tokens quadratically in the future while maintaining high fidelity. Our method achieves significant speedups through supervised fine-tuning on pretrained models. For example, it generates code and math nearly 5x faster, and improves general chat and knowledge tasks by almost 2.5x. These gains come without any loss in quality.
format Preprint
id arxiv_https___arxiv_org_abs_2507_11851
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Your LLM Knows the Future: Uncovering Its Multi-Token Prediction Potential
Samragh, Mohammad
Kundu, Arnav
Harrison, David
Nishu, Kumari
Naik, Devang
Cho, Minsik
Farajtabar, Mehrdad
Computation and Language
Machine Learning
Autoregressive language models are constrained by their inherently sequential nature, generating one token at a time. This paradigm limits inference speed and parallelism, especially during later stages of generation when the direction and semantics of text are relatively certain. In this work, we propose a novel framework that leverages the inherent knowledge of vanilla autoregressive language models about future tokens, combining techniques to realize this potential and enable simultaneous prediction of multiple subsequent tokens. Our approach introduces several key innovations: (1) a masked-input formulation where multiple future tokens are jointly predicted from a common prefix; (2) a gated LoRA formulation that preserves the original LLM's functionality, while equipping it for multi-token prediction; (3) a lightweight, learnable sampler module that generates coherent sequences from the predicted future tokens; (4) a set of auxiliary training losses, including a consistency loss, to enhance the coherence and accuracy of jointly generated tokens; and (5) a speculative generation strategy that expands tokens quadratically in the future while maintaining high fidelity. Our method achieves significant speedups through supervised fine-tuning on pretrained models. For example, it generates code and math nearly 5x faster, and improves general chat and knowledge tasks by almost 2.5x. These gains come without any loss in quality.
title Your LLM Knows the Future: Uncovering Its Multi-Token Prediction Potential
topic Computation and Language
Machine Learning
url https://arxiv.org/abs/2507.11851