首页 > AI前沿 > Diffusion Tree Sampling for Inference Time Adaptation of Crystal Generative Models

Diffusion Tree Sampling for Inference Time Adaptation of Crystal Generative Models

arXiv机器学习 2026-11-13 03:40 9 阅读 查看原文
用户要求将一段纯文本的新闻正文排版成语义化的HTML片段。 规则: 1. 只输出HTML片段本身,不要Markdown围栏、不要解释。 2. 每个自然段落用`

`包裹,长文用`

`/`

`划分小节。 3. 列表内容用`
  • `,引用用`
    `,代码用`
    `,重要句子可用``。
    4. 保留原文所有信息,不得增删改内容。
    5. 不要输出``/``/``等文档级标签。
    
    正文内容分析:
    这段正文是一个学术摘要/新闻稿,分为几个自然段落(其实原文只有一段长文本,但可以根据逻辑划分为几个部分,或者就保持一个段落。不过规则说“每个自然段落用`

    `包裹,长文用`

    `/`

    `划分小节”。原文没有明显的小节标题,我可以加一些语义化的`

    `或`

    `来划分,或者直接把整段作为`

    `。为了体现“长文用`

    `/`

    `划分小节”,我可以把这段文本分为几个逻辑小节,比如“背景与方法”、“技术贡献”、“实验与结果”。但是规则4要求“不得增删改内容”,如果加了小标题,算不算“增”内容? “不得增删改内容”通常指不改变原文的字句。如果加小标题,可能会被视为增加内容。为了安全起见,最好不加小标题,或者把原文本中的句子用`

    `分开,或者直接整个作为`

    `。 等等,原文是一段完整的文本,没有换行。如果强行拆成几个`

    `,也是排版需要。 让我们看原文: "Diffusion-based generative models have become strong priors over plausible crystal structures, but turning that prior into promising candidates requires steering toward several conflicting properties at once. Fine-tuning is expensive and produces a new model for every new objective, while differentiable guidance is not applicable to many real-world tasks. We instead adapt these generative models at inference time using Diffusion Tree Sampling (DTS), a Monte Carlo tree search procedure over the denoising trajectory. We make two technical contributions on top of DTS that are particularly relevant for AI-for-science applications. First, we show how to use existing property-conditional generative models (e.g., models fine-tuned for classifier-free guidance) as guided proposals inside DTS, without distorting the target distribution. Second, we extend DTS to multi-objective sampling, so that a single search produces samples spanning an entire Pareto front. We instantiate both adaptations on de novo crystal generation: improving the stability of crystals generated by different models, using pretrained conditional models to generate crystals belonging to rare space groups or with high bulk modulus, and generating crystals along a Pareto front of high dielectric constant and high bandgap materials. Across these tasks, DTS-based inference-time adaptation successfully improves model outputs without modifying the models themselves." 既然不能增删改内容,我就不添加`

    `/`

    `标题,因为添加标题等于增加了原文没有的文本。我可以将原文按句子或逻辑块拆分为几个`

    `,或者就整体用一个`

    `包裹。 为了体现排版,我可以把前半部分(背景和方法)作为第一个`

    `,后半部分(贡献和应用)作为第二个`

    `。或者就整个作为`

    `。 “每个自然段落用`

    `包裹”,原文在纯文本中就是一个自然段落。所以用一个`

    `包裹是符合规则的。 或者,我可以把句子拆分,但“自然段落”通常指原文中的段落。原文只有一段。 再检查规则3:列表内容用`

    • `,引用用`
      `,代码用`
      `,重要句子可用``。
      原文中没有明显的列表、引用、代码。我可以给一些重要的句子加上``。
      比如:
      `We instead adapt these generative models at inference time using Diffusion Tree Sampling (DTS), a Monte Carlo tree search procedure over the denoising trajectory.`
      `We make two technical contributions on top of DTS that are particularly relevant for AI-for-science applications.`
      `Across these tasks, DTS-based inference-time adaptation successfully improves model outputs without modifying the models themselves.`
      
      这样既使用了``,又没有改变原文内容。
      
      段落划分:
      原文:
      Diffusion-based generative models have become strong priors over plausible crystal structures, but turning that prior into promising candidates requires steering toward several conflicting properties at once. Fine-tuning is expensive and produces a new model for every new objective, while differentiable guidance is not applicable to many real-world tasks. We instead adapt these generative models at inference time using Diffusion Tree Sampling (DTS), a Monte Carlo tree search procedure over the denoising trajectory. We make two technical contributions on top of DTS that are particularly relevant for AI-for-science applications. First, we show how to use existing property-conditional generative models (e.g., models fine-tuned for classifier-free guidance) as guided proposals inside DTS, without distorting the target distribution. Second, we extend DTS to multi-objective sampling, so that a single search produces samples spanning an entire Pareto front. We instantiate both adaptations on de novo crystal generation: improving the stability of crystals generated by different models, using pretrained conditional models to generate crystals belonging to rare space groups or with high bulk modulus, and generating crystals along a Pareto front of high dielectric constant and high bandgap materials. Across these tasks, DTS-based inference-time adaptation successfully improves model outputs without modifying the models themselves.
      
      可以分成两个`

      `,逻辑更清晰:

      Diffusion-based generative models have become strong priors over plausible crystal structures, but turning that prior into promising candidates requires steering toward several conflicting properties at once. Fine-tuning is expensive and produces a new model for every new objective, while differentiable guidance is not applicable to many real-world tasks. We instead adapt these generative models at inference time using Diffusion Tree Sampling (DTS), a Monte Carlo tree search procedure over the denoising trajectory.

      We make two technical contributions on top of DTS that are particularly relevant for AI-for-science applications. First, we show how to use existing property-conditional generative models (e.g., models fine-tuned for classifier-free guidance) as guided proposals inside DTS, without distorting the target distribution. Second, we extend DTS to multi-objective sampling, so that a single search produces samples spanning an entire Pareto front. We instantiate both adaptations on de novo crystal generation: improving the stability of crystals generated by different models, using pretrained conditional models to generate crystals belonging to rare space groups or with high bulk modulus, and generating crystals along a Pareto front of high dielectric constant and high bandgap materials. Across these tasks, DTS-based inference-time adaptation successfully improves model outputs without modifying the models themselves.

      检查是否有增删改内容: 原文一字未改,只是加了HTML标签和``。 没有加`

      `/`

      `,因为加标题会引入新文本,违反“不得增删改内容”的严格解释。虽然规则说“长文用`

      `/`

      `划分小节”,但如果原文没有小节标题,硬加就是增内容。这里只有一段,也不算“长文”(大概300字)。所以不用标题是合理的。如果非要用,可以用`

      `包裹原文中的某句话作为小节名?不行,那会打乱句子结构。所以省略`

      `/`

      `是安全的。 再检查规则: 1. 只输出HTML片段本身,不要Markdown围栏、不要解释。 -> 好的。 2. 每个自然段落用`

      `包裹,长文用`

      `/`

      `划分小节。 -> 分为两个`

      `。 3. 列表内容用`

      • `,引用用`
        `,代码用`
        `,重要句子可用``。 -> 使用了``。
        4. 保留原文所有信息