Conflict-Aware Additive Guidance for Flow Models under Compositional Rewards
Quick Answer
The paper introduces Conflict-Aware Additive Guidance ($g^{car}$), a method that mitigates off-manifold drift in flow models during guided sampling.
Quick Take
The paper introduces Conflict-Aware Additive Guidance ($g^{car}$), a method that mitigates off-manifold drift in flow models during guided sampling. By dynamically resolving gradient conflicts, $g^{car}$ enhances generation fidelity across diverse applications, outperforming existing methods while maintaining low computational costs.
Key Points
- Conflict-Aware Additive Guidance ($g^{car}$) actively resolves gradient conflicts.
- The method improves generation fidelity in flow models without fine-tuning.
- Validation across synthetic datasets, image editing, and decision-making tasks.
- Outperforms baseline models while using light computational resources.
- Code for $g^{car}$ is publicly available for further research.
Paper Resources
📖 Reader Mode
~2 min readAbstract:Inference-time guided sampling steers state-of-the-art diffusion and flow models without fine-tuning by interpreting the generation process as a controllable trajectory. This provides a simple and flexible way to inject external constraints (e.g., cost functions or pre-trained verifiers) for controlled generation. However, existing methods often fail when composing multiple constraints simultaneously, which leads to deviations from the true data manifold. In this work, we identify root causes of this off-manifold drift and find that the approximation error scales severely with gradient misalignment. Building on these findings, we propose Conflict-Aware Additive Guidance ($g^\text{car}$), a lightweight and learnable method, which actively rectifies off-manifold drift by dynamically detecting and resolving gradient conflicts. We validate $g^\text{car}$ across diverse domains, ranging from synthetic datasets and image editing to generative decision-making for planning and control. Our results demonstrate that $g^\text{car}$ effectively rectifies off-manifold drift, surpassing baselines in generation fidelity while using light compute. Code is available at this https URL.
| Comments: | Forty-Third International Conference on Machine Learning (ICML 2026) |
| Subjects: | Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG); Robotics (cs.RO) |
| Cite as: | arXiv:2605.20758 [cs.AI] |
| (or arXiv:2605.20758v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2605.20758 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Xuehui Yu [view email]
[v1]
Wed, 20 May 2026 05:56:55 UTC (11,025 KB)
— Originally published at arxiv.org
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.AI
See more →Procedural Memory Distillation: Online Reflection for Self-Improving Language Models
Procedural Memory Distillation (PMD) enhances reinforcement learning by converting cross-episode signals into reusable memory, improving Qwen3-8B and OLMo3-Instruct-7B models by 3.8-5.5% on SCIKNOWEVAL and 7.9-13.6% on . The co-evolution of policy and memory allows for more effective self-supervision, demonstrating significant performance gains when both components are active.