Continuous Visual Autoregressive Generation via Score Maximization

TL;DR

This paper introduces a novel continuous visual autoregressive model that directly generates visual data without quantization, using score maximization based on proper scoring rules, especially the energy score, to improve generative quality.

Contribution

The paper proposes a continuous VAR framework leveraging proper scoring rules, enabling direct visual data generation without quantization, which is a significant advancement over traditional methods.

Findings

Framework based on proper scoring rules improves generation quality.

Energy score enables likelihood-free training in continuous space.

Connections established with previous methods like GIVT and diffusion loss.

Abstract

Conventional wisdom suggests that autoregressive models are used to process discrete data. When applied to continuous modalities such as visual data, Visual AutoRegressive modeling (VAR) typically resorts to quantization-based approaches to cast the data into a discrete space, which can introduce significant information loss. To tackle this issue, we introduce a Continuous VAR framework that enables direct visual autoregressive generation without vector quantization. The underlying theoretical foundation is strictly proper scoring rules, which provide powerful statistical tools capable of evaluating how well a generative model approximates the true distribution. Within this framework, all we need is to select a strictly proper score and set it as the training objective to optimize. We primarily explore a class of training objectives based on the energy score, which is likelihood-free and thus overcomes the difficulty of making probabilistic predictions in the continuous space. Previous efforts on continuous autoregressive generation, such as GIVT and diffusion loss, can also be derived from our framework using other strictly proper scores. Source code: https://github.com/shaochenze/EAR.