---
title: "TDLM: Adaptive Temporal Integration for Diffusion Language Models"
source: "arXiv AI/CL/LG（论文）"
source_tier: "t2"
url: "https://arxiv.org/abs/2610.36940v1"
permalink: "https://feed.astralor.com/items/itm_a8b30b84dc4b0eb36cb79c8b"
published: "2026-10-27T10:01:08.000Z"
score: 54
channel: "tech"
tags: ["TDLM","Adaptive Temporal Integration","Diffusion Language Models","Masked","Temporal Self-Con","Temporal Self-Conditioning Diffusion","Language Model"]
---

# TDLM: Adaptive Temporal Integration for Diffusion Language Models

## 速览

Masked diffusion language models repeatedly compute representations for unresolved tokens, but conventional denoisers do not explicitly retain these representations for reuse across steps. We introduce Temporal Self-Con…

## 关键点

- Masked diffusion language models repeatedly compute representations for unresolved tokens, but conventional denoisers do not explicitly retain these representations for reuse across steps.
- We introduce Temporal Self-Conditioning Diffusion Language Model (TDLM), which lets each unresolved position attend to its own history of hidden representations within a frozen denoiser.
- This gives later predictions access to computation already performed.

## 全文

Masked diffusion language models repeatedly compute representations for unresolved tokens, but conventional denoisers do not explicitly retain these representations for reuse across steps. We introduce Temporal Self-Conditioning Diffusion Language Model (TDLM), which lets each unresolved position attend to its own history of hidden representations within a frozen denoiser. This gives later predictions access to computation already performed. Training uses self-generated trajectories and a simple cross-entropy objective for tokens revealed between consecutively sampled decoding steps, while causal temporal attention permits parallel evaluation during training. On LLaDA, the temporal modules add only approximately 0.84% trainable parameters. Across LLaDA and Dream, the main comparison shows up to 7.15$\times$ faster generation than the corresponding original diffusion language model, with competitive answer quality on the evaluated tasks. These results motivate temporal memory as a practical way to make more effective use of repeated denoising computation.

---

引用前请核对原文;本文由星际流动整理:https://feed.astralor.com/items/itm_a8b30b84dc4b0eb36cb79c8b