Time-Annealed Perturbation Sampling: Diverse Generation for Diffusion Language Models

TL;DR

This paper introduces TAPS, a novel inference method for diffusion language models that enhances generation diversity by leveraging the temporal structure of the diffusion process, without additional training.

Contribution

It reveals the temporal division of labor in Diffusion-LMs and proposes TAPS, a training-free strategy to improve diversity while maintaining quality across various tasks.

Findings

TAPS improves diversity in diffusion language model outputs.

TAPS maintains fluency and instruction adherence.

Applicable to multiple diffusion model architectures.

Abstract

Diffusion language models (Diffusion-LMs) introduce an explicit temporal dimension into text generation, yet how this structure can be leveraged to control generation diversity for exploring multiple valid semantic or reasoning paths remains underexplored. In this paper, we show that Diffusion-LMs, like diffusion models in image generation, exhibit a temporal division of labor: early denoising steps largely determine the global semantic structure, while later steps focus on local lexical refinement. Building on this insight, we propose Time-Annealed Perturbation Sampling (TAPS), a training-free inference strategy that encourages semantic branching early in the diffusion process while progressively reducing perturbations to preserve fluency and instruction adherence. TAPS is compatible with both non-autoregressive and semi-autoregressive Diffusion backbones, demonstrated on LLaDA and TraDo in our paper, and consistently improves output diversity across creative writing and reasoning benchmarks without compromising generation quality.