Token Pruning for In-Context Generation in Diffusion Transformers
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2026
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866910008429510656 |
|---|---|
| author | Lin, Junqing Zheng, Xingyu Cheng, Pei Fu, Bin Sun, Jingwei Sun, Guangzhong |
| author_facet | Lin, Junqing Zheng, Xingyu Cheng, Pei Fu, Bin Sun, Jingwei Sun, Guangzhong |
| contents | In-context generation significantly enhances Diffusion Transformers (DiTs) by enabling controllable image-to-image generation through reference examples. However, the resulting input concatenation drastically increases sequence length, creating a substantial computational bottleneck. Existing token reduction techniques, primarily tailored for text-to-image synthesis, fall short in this paradigm as they apply uniform reduction strategies, overlooking the inherent role asymmetry between reference contexts and target latents across spatial, temporal, and functional dimensions. To bridge this gap, we introduce ToPi, a training-free token pruning framework tailored for in-context generation in DiTs. Specifically, ToPi utilizes offline calibration-driven sensitivity analysis to identify pivotal attention layers, serving as a robust proxy for redundancy estimation. Leveraging these layers, we derive a novel influence metric to quantify the contribution of each context token for selective pruning, coupled with a temporal update strategy that adapts to the evolving diffusion trajectory. Empirical evaluations demonstrate that ToPi can achieve over 30\% speedup in inference while maintaining structural fidelity and visual consistency across complex image generation tasks. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2602_01609 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Token Pruning for In-Context Generation in Diffusion Transformers Lin, Junqing Zheng, Xingyu Cheng, Pei Fu, Bin Sun, Jingwei Sun, Guangzhong Computer Vision and Pattern Recognition In-context generation significantly enhances Diffusion Transformers (DiTs) by enabling controllable image-to-image generation through reference examples. However, the resulting input concatenation drastically increases sequence length, creating a substantial computational bottleneck. Existing token reduction techniques, primarily tailored for text-to-image synthesis, fall short in this paradigm as they apply uniform reduction strategies, overlooking the inherent role asymmetry between reference contexts and target latents across spatial, temporal, and functional dimensions. To bridge this gap, we introduce ToPi, a training-free token pruning framework tailored for in-context generation in DiTs. Specifically, ToPi utilizes offline calibration-driven sensitivity analysis to identify pivotal attention layers, serving as a robust proxy for redundancy estimation. Leveraging these layers, we derive a novel influence metric to quantify the contribution of each context token for selective pruning, coupled with a temporal update strategy that adapts to the evolving diffusion trajectory. Empirical evaluations demonstrate that ToPi can achieve over 30\% speedup in inference while maintaining structural fidelity and visual consistency across complex image generation tasks. |
| title | Token Pruning for In-Context Generation in Diffusion Transformers |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2602.01609 |