TrajCFG is introduced, a training-free method for flow-matching models that adapts token-wise guidance using historical changes in guided velocity direction, and results support historical directional feedback as a broadly applicable signal for fine-grained guidance allocation.
Abstract
Classifier-free guidance (CFG) typically applies a shared scale to all latent tokens, although their velocity predictions evolve differently during sampling. We introduce Trajectory-Adaptive Classifier-Free Guidance (TrajCFG), a training-free method for flow-matching models that adapts token-wise guidance using historical changes in guided velocity direction. TrajCFG maps this feedback to bounded scales, assigning stronger guidance to larger directional changes and weaker guidance to more consistent directions. It combines adjacent-step and interval-normalized longer-span feedback, reusing standard CFG predictions without additional model evaluations. Controlled diagnostics and ablations examine the roles of directional feedback, token-wise allocation, and history-span fusion. Experiments on text-to-video, image-to-video, text-to-audio-video, text-to-image, and image-to-3D generation show improvements over base samplers across multiple backbones. For text-to-video generation, TrajCFG improves VBench Total scores by 0.30-0.75 points over the strongest evaluated baseline on each of three backbones. These results support historical directional feedback as a broadly applicable signal for fine-grained guidance allocation. The project page is available at https://jyliu-98.github.io/trajcfg-project-page/.
It is proved this decoupled objective is mathematically exact: the centered advantage cancels the anchor in expectation, filtering out the reward-neutral guidance offset and retaining only the reward-correlated covariance increment, yet inference-time CFG preserves its original semantics.
Zhe-Ming Fu, Wen-Bin Ning, R.-Z. He et al.· 1 citation
Classifier-free guidance (CFG) improves conditional generation in Flow Matching, but strong guidance can distort the generated distribution and reduce diversity. We provide a geometric account of this behavior by viewing Flow Matching as a time-varying gradient flow and characterizing how CFG reshapes its underlying po...
Rectified flow models are increasingly used for text-to-image generation, a core multimedia content generation task, thanks to efficient, near-deterministic ODE sampling, but aligning them with human preferences or prompt-following objectives still commonly relies on reinforcement learning with stochastic rollouts. Exi...
Shuo Zhang, Qi-Shan Zhang, Hua-Yu Zhang et al.· Proceedings of the 34th ACM...· 0 citations
Directly aligning trajectory-distilled video generators slashes rollout costs and optimizes the deployed few-step model without teacher fine-tuning or re-distillation. However, conventional flow-matching preference objectives assume single-head local velocity supervision, mismatching the finite-interval, block-average...
Jing-Yuan Zhu, Biao-Long Chen, Yang-Yang Qin et al.· 0 citations
Explicit point correspondence is attractive for adapting video generators: it describes how visible surfaces move, but can overwrite the appearance prior learned by a large pretrained model. We study this tension in a 5.01B-parameter text-to-video diffusion transformer. ACROSSVAM2.0 combines a confidence-weighted track...
The practical success of conditional image generation hinges on fine-grained differences in condition alignment and visual fidelity. Classifier-free guidance (CFG) is central to this success, but its lack of an explicit criterion makes it difficult to assess whether the guided trajectory is progressing as intended. To...
Gihoon Kim, Taesup Kim· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.