---
id: 20260718-T0-05
title: "连续扩散语言模型TTCD：无需采样，一步生成完整Token"
title_en: "TTCD Diffusion LM Generates Complete Token Canvas in One Step Without Sampling"
url: https://ai.daily.yangsir.net/daily/20260718-T0-05
issue_date: 2026-07-18
publish_date: 2026-07-17T04:00:00.000Z
category: research
source_name: "arXiv cs.CL (NLP)"
source_url: https://arxiv.org/abs/2607.14106
---

# 连续扩散语言模型TTCD：无需采样，一步生成完整Token

研究团队提出Token Time Continuous Diffusion (TTCD)，一种新的扩散语言模型。TTCD在连续空间中运行，将高斯噪声确定性映射为最终Token画布，无需额外采样步骤。该方法显著简化了文本生成流程，实现了更高效的LLM推理，在保持生成质量的同时大幅减少计算开销。

## English Version

**TTCD Diffusion LM Generates Complete Token Canvas in One Step Without Sampling**

Researchers introduce Token Time Continuous Diffusion (TTCD), a novel diffusion language model that operates in continuous space by deterministically mapping Gaussian noise to a final token canvas without additional sampling. This approach dramatically simplifies the text generation pipeline, enabling more efficient LLM inference with reduced computational overhead while maintaining output quality.

---

**来源**：[arXiv cs.CL (NLP)](https://arxiv.org/abs/2607.14106)

**详情页**：https://ai.daily.yangsir.net/daily/20260718-T0-05

---

*智语观潮 · Daily — https://ai.daily.yangsir.net/llms.txt*