发表机构
University of Oxford; University of British Columbia; University of Cambridge（牛津大学; 不列颠哥伦比亚大学; 剑桥大学）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
提出ENCORE，首个精确并行推理时控制方法，通过存储轨迹避免模拟难解时间反转，实现无偏引导，适用于多种生成任务。
AI 中文摘要
推理时控制无需重新训练即可将预训练生成模型引导至目标分布。我们研究倾斜目标 $\pi_0\propto G_0\\,p_0$，其中 $p_0$ 是采样器输出分布，$G_0$ 是可评估的重新加权函数。现有方法依赖于使用序贯蒙特卡洛（SMC）的序贯退火或使用副本交换（RE）的并行退火。序贯控制是精确的，但需要大量粒子群体，而目前不存在精确的并行控制方法：现有的RE修正近似了一个难以处理的时间反转，因此存在偏差。我们提出了精确非平衡控制与副本交换（ENCORE），这是第一种精确的并行控制方法。每个副本存储其生成轨迹，因此向上移动是截断操作，且难以处理的时间反转从未被模拟。我们证明了目标不变性，并表明所得动力学是精确时间反转作为前向提议的非平衡副本交换动力学。在正则条件下，我们的扩散分析表明，在没有引导的情况下，序贯和并行控制在时间离散化细化时都会变得不稳定，而引导提议保持稳定，并为调整时间表和计算预算提供诊断。在合成目标、生物分子玻尔兹曼采样和图像生成中，ENCORE 实现了具有竞争力的准确性和多样性，对采样器扰动保持鲁棒性，并适用于现有RE修正不可用的蒸馏采样器。
英文摘要
Inference-time control steers a pretrained generative model towards a target distribution without retraining. We study tilted targets $π_0\propto G_0\,p_0$, where $p_0$ is the sampler output distribution and $G_0$ is an evaluable reweighting function. Existing approaches rely on sequential annealing with sequential Monte Carlo (SMC) or parallel annealing with replica exchange (RE). Sequential control is exact but needs large particle populations, whereas no exact parallel control method exists: existing RE corrections approximate an intractable time reversal and are biased. We propose Exact Non-equilibrium COntrol with Replica Exchange (ENCORE), the first exact parallel control method. Each replica stores its generation trajectory, so the upward move is a truncation and the intractable time reversal is never simulated. We prove target invariance and show that the resulting dynamics are those of non-equilibrium replica exchange with the exact time reversal as forward proposal. Under regularity conditions, our diffusion analysis shows that both sequential and parallel control become unstable under refinement of the time discretisation without guidance, whereas guided proposals remain stable and yield diagnostics for tuning the schedule and the computational budget. Across synthetic targets, Boltzmann sampling of biomolecules, and image generation, ENCORE achieves competitive accuracy and diversity, remains robust to sampler perturbations, and applies to distilled samplers where existing RE corrections are unavailable.
CommentsA shorter version of this work was accepted at the NeurIPS 2026 PriGM Workshop
