D-DOIT：通过Doob's h-变换实现离散扩散的无训练适应

中文摘要 AI 辅助

我们提出D-DOIT（离散Doob导向推理时变换），一种针对具有通用奖励的离散扩散模型的无训练且高效的适应方法。D-DOIT将适应问题表述为从奖励倾斜的目标分布中采样，并通过离散扩散反向核的Doob's h-变换实现这种传输，仅使用奖励值而非奖励梯度。与连续扩散不同，掩蔽离散扩散采样的是分类词元揭示转换，而非连续状态更新。D-DOIT推导了相应的离散Doob's h-变换，通过重新加权反向转换概率而非添加漂移校正来引导采样。为使该变换实用化，D-DOIT避免了昂贵的未来轨迹展开。在每个引导步骤中，D-DOIT采样候选下一状态，使用模型预测头将每个候选补全为干净序列，用奖励预言机评估每个补全，并按与奖励成比例的概率重新采样下一状态。可选的后期最佳K细化通过仅在去噪接近结束时分支轨迹进一步提高样本质量，避免了整个轨迹上的K倍成本。实验上，在调控DNA设计和蛋白质逆折叠基准上，D-DOIT优于无训练引导基线。它提高了增强子活性和细胞类型特异性，同时保持序列自然性，并在蛋白质逆折叠中取得了最高成功率。

英文摘要

We propose D-DOIT (Discrete Doob-Oriented Inference-time Transformation), a training-free and efficient adaptation method for discrete diffusion models with generic rewards. D-DOIT formulates adaptation as sampling from a reward-tilted target distribution and realizes this transport through Doob's h-transform of the discrete diffusion reverse kernel, using only reward values rather than reward gradients. Unlike continuous diffusion, masked discrete diffusion samples categorical token-reveal transitions rather than continuous state updates. D-DOIT derives the corresponding discrete Doob's h-transform, which guides sampling by reweighting reverse transition probabilities instead of adding a drift correction. To make this transformation practical, D-DOIT avoids expensive future rollouts. At each guided step, D-DOIT samples candidate next states, uses the model prediction head to complete each candidate into a clean sequence, evaluates each completion with the reward oracle, and resamples the next state with probabilities proportional to the rewards. An optional late-stage best-of-K refinement further improves sample quality by branching trajectories only near the end of denoising, avoiding the $K$-fold cost over the full trajectory. Empirically, across regulatory DNA design and protein inverse folding benchmarks, D-DOIT outperforms training-free guidance baselines. It improves enhancer activity and cell-type specificity while preserving sequence naturalness, and achieves the highest success rate in protein inverse folding.

发表机构

• Northwestern University（西北大学）
• King Abdullah University of Science and Technology（阿卜杜拉国王科技大学）

机构由 AI 辅助整理，请以论文原文为准。