可解释提示词变得易于编辑：面向编辑友好硬提示词反演的 dLLM 词元间相似度抑制
Interpretable Prompts made Edit-Friendly: Token-to-Token Similarity Reduction in dLLMs for Edit-Friendly Hard Prompt Inversion
把扩散语言模型解码器中的词元相似度控制与扩散时间采样结合，得到对词元替换/追加鲁棒的硬提示词，可直接用于 T2I 模型的可编辑提示词反演与下游提示词编辑流程，且无需微调 T2I 或语言模型。
原文依据 By coupling diffusion-time sampling with token-similarity control inside a dLLM decoder
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论通过提示词工程构造提示词，使模型的内部表征朝向特定且预定义的结果，这一过程可能十分耗时，且往往需要多次迭代。硬提示词反演提供了一种互补的工作流：从参考图像出发，生成一个提示词，使文本到图像（T2I）模型以该提示词为条件重建参考图像。现有的反演方法要么产生不连贯的文本，要么产生的提示词对下游的词元编辑过于敏感。我们提出了一种基于 dLLM 的提示词反演框架，其产生的提示词（i）对人类更易解释，（ii）与参考图像对齐更好，（iii）面向下游的词元替换与词元追加操作（即编辑友好的提示词）。该方法是即插即用的，既不需要微调 T2I 模型，也不需要微调 dLLM。在三个数据集上的实验表明，相对于现有的提示词反演基线，反演时间约减少 10 倍，可解释性得分更高，并且以 TIFA、GPT-V 偏好评分和受控用户研究衡量的提示词可编辑性显著更高，同时保持高保真的图像生成。通过在 dLLM 解码器内部将扩散时间采样与词元相似度控制相结合，我们的方法将提示词反演从重建扩展到下游的词元编辑任务，从而获得更快、更具可迁移性、并能泛化到多个 T2I 模型的提示词。
英文原文摘要 1451 字符
Crafting prompts via Prompt Engineering that steer a model's internal representations toward specific and pre-defined outcomes can be time-consuming, often requiring multiple iterations. Hard Prompt Inversion offers a complementary workflow: start from a reference image and generate a prompt that conditions a text-to-image (T2I) model to reconstruct the reference image. Existing inversion methods either yield incoherent text, or produce prompts that are overly sensitive to downstream token edits. We propose a dLLM-based prompt inversion framework that yield prompts that are (i) more interpretable to humans, (ii) better aligned with the reference image, and (iii) designed for downstream token swap and token append operations (aka edit-friendly prompts). The method is plug-and-play, requiring no finetuning of either the T2I model or the dLLM. Experiments across three datasets show a ~10xreduction in inversion time relative to existing prompt-inversion baselines, higher interpretability scores, and significantly higher prompt editability, as measured by TIFA, GPT-V preference scoring, and controlled user studies, all while preserving high-fidelity image generation. By coupling diffusion-time sampling with token-similarity control inside a dLLM decoder, our approach extends prompt inversion beyond reconstruction to downstream token-editing tasks, enabling faster, more transferable prompts that generalize across multiple T2I models.