arXiv cs.CL· Ryandito Diandaru, Ahmed Elshabrawy, Fadli Aulawi Al Ghiffari, Ikhlasul Akmal Hanif, Alham Fikri Aji·· 12 小时前AI 评分24
TimpaTeks：用扩散语言模型引导实现文本原地修改
TimpaTeks: Automatic In-place Text Sequence Modification via Diffusion Language Model Steering
AI 导读
TimpaTeks 将激活引导扩展到扩散语言模型，实现无需指令微调模型的文本原地修改，在 IMDB 影评情感和合成猫狗数据集的概念引导实验中验证可行。该方法在修改的同时降低句子困惑度并保留原句结构，且因原地去噪而非构造额外提示词条件输出序列，计算成本低于基于提示词的 DLM 引导。
正文 · 原文
Abstract:We extend activation steering to diffusion language models (DLMs) and study a novel problem that arose due to the inference mechanism of DLMs: Modifying a text in-place to manifest a different concept. We propose TimpaTeks, an automatic in-place text modification mechanism using DLMs. Experiments on IMDB movie reviews (sentiment) and a synthetic Cats and Dogs Dataset (arbitrary, more unconventional concept steering) show that TimpaTeks provides a feasible novel mechanism to steer diffusion language model outputs in-place. TimpaTeks enables in-place modification while simultaneously lowers sentence perplexity and retaining the original sentence structre without the need of instruction tuned models. TimpaTeks is also computationally cheaper than prompt-based DLM steering, as it performs denoising in-place rather than constructing an additional prompt-conditioned output sequence.
| Comments: | 16 pages |
| Subjects: | Computation and Language (cs.CL); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2606.08408 [cs.CL] |
| (or arXiv:2606.08408v2 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2606.08408 arXiv-issued DOI via DataCite |
Submission history
From: Ikhlasul Akmal Hanif [view email]
[v1]
Sun, 7 Jun 2026 02:13:17 UTC (1,871 KB)
[v2]
Fri, 16 Oct 2026 03:30:18 UTC (1,165 KB)
来源：arXiv cs.CL · arxiv.org