首页 > AI前沿 > High-Dimensional Black-Box Optimization via Guidance Sampling of Diffusion Models

High-Dimensional Black-Box Optimization via Guidance Sampling of Diffusion Models

arXiv机器学习 2026-11-17 23:51 9 阅读 查看原文
用户需要我作为专业的新闻排版工程师,将提供的纯文本新闻正文排版成语义化的HTML片段。 需要遵循的规则: 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。
    
    提供的正文:
    Many real-world optimization problems involve optimizing black-box objectives for which only a limited number of evaluations are available. Bayesian optimization (BO) is a powerful framework for such problems, but its effectiveness often deteriorates in high-dimensional spaces. Recent work leverages diffusion models for black-box optimization, often through pre-training a reward-conditioned score network to steer generation towards high-reward samples. Alternatively, gradient-based guidance methods assume access to a differentiable reward function from which guidance gradients are queried. However, these approaches either require large amounts of labeled data for offline training or access to exact reward gradients, limiting their applicability in black-box settings. We propose BReaD (Black-box Reward-guided Sampling of Diffusion Model), a guidance sampling framework for optimizing black-box rewards without assuming access to a labeled dataset or reward gradients. The proposed method uses a Gaussian process (GP) as a differentiable surrogate for the unknown objective and actively queries labels for the generated samples. We provide a theoretical analysis of the total-variation distance between the generated distribution and the target distribution. Using Upper Confidence Bound (UCB) as guidance signals, our method balances exploration and exploitation during generation and enables the discovery of high-reward regions in the data space. Numerical experiments on scientific discovery and image generation tasks demonstrate that our framework can effectively optimize black-box rewards in low-data scenarios.
    
    分析正文结构:
    这段正文主要介绍了一个名为BReaD的框架,用于黑盒奖励优化。可以将其分为几个小节,或者直接作为几个段落。
    为了体现“长文用

    /

    划分小节”,我可以将这段文本划分为几个逻辑小节。 比如: - 背景与挑战 - 方法提出 - 理论与实验 不过原文其实是一整段连贯的学术摘要。如果强行划分小节,需要添加标题,但规则4说“不得增删改内容”。如果我添加了`

    `标题,算不算“增”内容? 规则2说“长文用

    /

    划分小节”,这意味着可以添加小节标题来组织结构。但如果不想添加额外文字,也可以把原文按句子或逻辑拆分成多个`

    `,或者用`

    `作为小节标题(比如用原文中的关键词作为标题,或者直接不添加标题,仅用`

    `)。 为了安全起见,避免添加原文没有的文字作为标题,我可以将其拆分为几个`

    `段落,或者使用`

    `但标题内容必须来自原文(这很难)。 其实,规则2说“长文用

    /

    划分小节”,如果不添加标题,直接全用`

    `也是可以的,或者我可以用原文中的句子作为小节标题?不,那样会打乱原文。 最好的做法是:将原文按逻辑拆分为几个`

    `,并在适当的位置使用``来强调重要句子(如提出方法的句子)。因为原文本身只是一段摘要,没有明显的列表、引用或代码。 让我们把原文拆分成几个段落: 段落1:背景和现有方法。 Many real-world optimization problems involve optimizing black-box objectives for which only a limited number of evaluations are available. Bayesian optimization (BO) is a powerful framework for such problems, but its effectiveness often deteriorates in high-dimensional spaces. Recent work leverages diffusion models for black-box optimization, often through pre-training a reward-conditioned score network to steer generation towards high-reward samples. Alternatively, gradient-based guidance methods assume access to a differentiable reward function from which guidance gradients are queried. However, these approaches either require large amounts of labeled data for offline training or access to exact reward gradients, limiting their applicability in black-box settings. 段落2:提出的方法。 We propose BReaD (Black-box Reward-guided Sampling of Diffusion Model), a guidance sampling framework for optimizing black-box rewards without assuming access to a labeled dataset or reward gradients. The proposed method uses a Gaussian process (GP) as a differentiable surrogate for the unknown objective and actively queries labels for the generated samples. 段落3:理论和实验。 We provide a theoretical analysis of the total-variation distance between the generated distribution and the target distribution. Using Upper Confidence Bound (UCB) as guidance signals, our method balances exploration and exploitation during generation and enables the discovery of high-reward regions in the data space. Numerical experiments on scientific discovery and image generation tasks demonstrate that our framework can effectively optimize black-box rewards in low-data scenarios. 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。 -> 这里没有特别长的长文,拆分成3个`

    `即可。如果非要用`

    `,我可以用原文中的概念作为标题,但这可能被视为“增”内容。为了严格遵守“不得增删改内容”,不添加原文没有的标题文字,只用`

    `包裹,并在重要句子用``。或者,我可以把第一句话作为`

    `?不,那会改变原文结构。直接用`

    `最安全。 3. 列表、引用、代码按需使用,这里没有。重要句子用``。 -> 已使用。 4. 保留原文所有信息,不得增删改内容。 -> 确保单词、标点完全一致。 5. 不要输出文档级标签。 -> 好的。 仔细核对原文和HTML转换,确保没有拼写错误或遗漏。 原文: Many real-world optimization problems involve optimizing black-box objectives for which only a limited number of evaluations are available. Bayesian optimization (BO) is a powerful framework for such problems, but its effectiveness often deteriorates in high-dimensional spaces. Recent work leverages diffusion models for black-box optimization, often through pre-training a reward-conditioned score network to steer generation towards high-reward samples. Alternatively, gradient-based guidance methods assume access to a differentiable reward function from which guidance gradients are queried. However, these approaches either require large amounts of labeled data for offline training or access to exact reward gradients, limiting their applicability in black-box settings. We propose BReaD (Black-box Reward-guided Sampling of Diffusion Model), a guidance sampling framework for optimizing black-box rewards without assuming access to a labeled dataset or reward gradients. The proposed method uses a Gaussian process (GP) as a differentiable surrogate for the unknown objective and actively queries labels for the generated samples. We provide a theoretical analysis of the total-variation distance between the generated distribution and the target distribution. Using Upper Confidence Bound (UCB) as guidance signals, our method balances exploration and exploitation during generation and enables the discovery of high-reward regions in the data space. Numerical experiments on scientific discovery and image generation tasks demonstrate that our framework can effectively optimize black-box rewards in low-data scenarios. 拆分段落: P1: Many real-world optimization problems involve optimizing black-box objectives for which only a limited number of evaluations are available. Bayesian optimization (BO) is a powerful framework for such problems, but its effectiveness often deteriorates in high-dimensional spaces. Recent work leverages diffusion models for black-box optimization, often through pre-training a reward-conditioned score network to steer generation towards high-reward samples. Alternatively, gradient-based guidance methods assume access to a differentiable reward function from which guidance gradients are queried. However, these approaches either require large amounts of labeled data for offline training or access to exact reward gradients, limiting their applicability in black-box settings. P2: We propose BReaD (Black-box Reward-guided Sampling of Diffusion Model), a guidance sampling framework for optimizing black-box rewards without assuming access to a labeled dataset or reward gradients. The proposed method uses a Gaussian process (GP) as a differentiable surrogate for the unknown objective and actively queries labels