TokenTrim prunes corrupted tokens during video generation inference to eliminate temporal drift.
Auto-regressive video generation enables long video synthesis by iteratively conditioning each new batch of frames on previously generated content. However, recent work has shown that such pipelines suffer from severe temporal drift, where errors accumulate and amplify over long horizons. We hypothesize that this drift does not primarily stem from insufficient model capacity, but rather from inference-time error propagation. Specifically, we contend that drift arises from the uncontrolled reuse of corrupted latent conditioning tokens during auto-regressive inference. To correct this accumulation of errors, we propose a simple, inference-time method that mitigates temporal drift by identifying and removing unstable latent tokens before they are reused for conditioning. For this purpose, we define unstable tokens as latent tokens whose representations deviate significantly from those of the previously generated batch, indicating potential corruption or semantic drift. By explicitly removing corrupted latent tokens from the auto-regressive context, rather than modifying entire spatial regions or model parameters, our method prevents unreliable latent information from influencing future generation steps. As a result, it significantly improves long-horizon temporal consistency without modifying the model architecture, training procedure, or leaving latent space.
TokenTrim removes corrupted tokens from cache during inference to reduce temporal drift in video generation.
The method improves video quality by 5.58-6.90% and reduces temporal drift by 26 percentage points.
Inference-time token pruning addresses error compounding without requiring model retraining or architectural changes.
The script only discusses results on Wan model; generalization to other video models unclear.
Method requires computing per-token drift scores at each step, potentially adding computational overhead.
Shenghai Yuan, Yuanyang Yin +4
We introduce Helios, the first 14B video generation model that runs at 19.5 FPS on a single NVIDIA H100 GPU and supports minute-scale generation while matching the quality of a strong baseline. We mak...
Yiying Yang, Wei Cheng +6
OmniLottie is a versatile framework that generates high quality vector animations from multi-modal instructions. For flexible motion and visual content control, we focus on Lottie, a light weight JSON...
Zengbin Wang, Xuecai Hu +4
Text-to-image (T2I) models have achieved remarkable success in generating high-fidelity images, but they often fail in handling complex spatial relationships, e.g., spatial perception, reasoning, or i...
Grigorii Alekseenko, Aleksandr Gordeev +8
Instruction-based image editing is among the fastest developing areas in generative AI. Over the past year, the field has reached a new level, with dozens of open-source models released alongside high...
Hojung Jung, Rodrigo Hormazabal +6
Molecular generation with diffusion models has emerged as a promising direction for AI-driven drug discovery and materials science. While graph diffusion models have been widely adopted due to the dis...
