FlashMesh: Faster and Better Autoregressive Mesh Synthesis via Structured Speculation
This addresses the bottleneck of slow decoding for interactive and large-scale 3D mesh applications, representing an incremental improvement by optimizing an existing method.
The paper tackles the slow inference problem of autoregressive 3D mesh generation by introducing FlashMesh, a framework that uses a predict-correct-verify paradigm and speculative decoding to achieve up to a 2x speedup while improving generation fidelity.
Autoregressive models can generate high-quality 3D meshes by sequentially producing vertices and faces, but their token-by-token decoding results in slow inference, limiting practical use in interactive and large-scale applications. We present FlashMesh, a fast and high-fidelity mesh generation framework that rethinks autoregressive decoding through a predict-correct-verify paradigm. The key insight is that mesh tokens exhibit strong structural and geometric correlations that enable confident multi-token speculation. FlashMesh leverages this by introducing a speculative decoding scheme tailored to the commonly used hourglass transformer architecture, enabling parallel prediction across face, point, and coordinate levels. Extensive experiments show that FlashMesh achieves up to a 2 x speedup over standard autoregressive models while also improving generation fidelity. Our results demonstrate that structural priors in mesh data can be systematically harnessed to accelerate and enhance autoregressive generation.