aiqbench analysis
The paper argues that porting diffusion-style guidance directly to flow-matching models breaks the coupling distribution used during training, which weakens control over sampling accuracy. Its proposed Coupled Guidance method uses that coupling structure and reports better sample quality and stability on robotics and image tasks at no extra computational cost. For ML practitioners, the main takeaway is a training-aware fix for a known instability, though the gains are reported on the authors' own benchmarks and need independent replication.
Coupled Guidance for Flow Matching
arXiv:2610.40109v1 Announce Type: new Abstract: Guidance methods are essential tools for steering the generation process in diffusion models, improving the generalizability of stochastic robotic controllers and enabling high-quality controlled image synthesis. Despite strong parallels between flow-matching and diffusion, integrating guidance into flow-matching models remains quite challenging. Recent work has proposed general guidance frameworks for flow-matching, however, they are direct adaptations of diffusion guidance methods and can prove unstable. In this work, we conduct a theoretical analysis of guidance in flow-matching and show that a naive translation of diffusion-based guidance to flow-matching breaks the coupling distribution used during training, resulting in weaker control over sampling accuracy. Building on this insight, we propose Coupled Guidance, a flow-matching-specific guidance mechanism derived from a geometric perspective that leverages the coupling structure inherent to flow-matching training. To validate the effectiveness of our approach, we conduct extensive experimentation on a custom robotic collision avoidance task using ManiSkill3, four robotic dexterous manipulation tasks from the Adroit benchmark, and three image generation tasks, namely inpainting, super-resolution, and deblurring, using the ImageNet-128 and CelebA-HQ-256 datasets. Our findings show that our framework improves sample quality and guidance stability across all tasks, without imposing any additional overhead.
Abstract:Guidance methods are essential tools for steering the generation process in diffusion models, improving the generalizability of stochastic robotic controllers and enabling high-quality controlled image synthesis. Despite strong parallels between flow-matching and diffusion, integrating guidance into flow-matching models remains quite challenging. Recent work has proposed general guidance frameworks for flow-matching, however, they are direct adaptations of diffusion guidance methods and can prove unstable. In this work, we conduct a theoretical analysis of guidance in flow-matching and show that a naive translation of diffusion-based guidance to flow-matching breaks the coupling distribution used during training, resulting in weaker control over sampling accuracy. Building on this insight, we propose Coupled Guidance, a flow-matching-specific guidance mechanism derived from a geometric perspective that leverages the coupling structure inherent to flow-matching training. To validate the effectiveness of our approach, we conduct extensive experimentation on a custom robotic collision avoidance task using ManiSkill3, four robotic dexterous manipulation tasks from the Adroit benchmark, and three image generation tasks, namely inpainting, super-resolution, and deblurring, using the ImageNet-128 and CelebA-HQ-256 datasets. Our findings show that our framework improves sample quality and guidance stability across all tasks, without imposing any additional overhead.
Accepted at NeurIPS 2026. 25 pages
Machine Learning (cs.LG); Robotics (cs.RO)
arXiv:2610.40109 [cs.LG]
(or arXiv:2610.40109v1 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2610.40109 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 291 words
Related coverage
StudentSim: Training LLM-based Student Simulators
arXiv · Sep 2, 2026
When Privileged Guidance Misaligns: State-Matched Routing and Contextualized Self-Distillation for Multi-Turn Agents
arXiv · Aug 7, 2026
Diffusion ReRoll: Revisable Denoising for Robotic Sequential Prediction
arXiv · Jul 23, 2026
Beyond End-to-End Success: Diagnosing Failures in Long-Horizon Security LLM Agents
arXiv · Aug 24, 2026