Diffusion vision-language models (DVLMs) process their entire input at every denoising step. Since visual tokens make up 89-97% of this input on the benchmarks we study, most of the decoding computati
LOWN adaptively skips image processing during diffusion vision-language model decoding when predictions agree, reducing computation while preserving accuracy.
Diffusion vision-language models (DVLMs) process their entire input at every denoising step. Since visual tokens make up 89-97% of this input on the benchmarks we study, most of the decoding computation is spent repeatedly processing the same image. Visual token pruning methods partially relieve this burden by removing part of the visual tokens, but they lose the removed information and still process the remaining image at every step. To measure how often the image is actually needed, we compare LLaDA-V's predictions with and without the image at every decoding step and find that removing the image changes a token being unmasked at only 22.8% of steps on average and at as few as 7.9%. We…
Summary and abstract excerpt via Hugging Face Papers ↗ · arXiv.
Look Only When Needed: Certificate-Guided Visual Routing for Efficient Diffusion VLMs was first seen on Hugging Face · Daily papers on Oct 30, 2026. Since then it has appeared in 14 checks on 2 different days. Its best position was #4 of 30, reached on Oct 30, 09:28 UTC. It is #5 right now, slipping from #4 to #5 over the last 6 checks.
Position on Hugging Face · Daily papers at every check (lower is better).
Best rank per day (UTC).
▲ as reported by Hugging Face.
Yes. As of Oct 31, 09:28 UTC it is #5, slipping from #4 to #5 over the last 6 checks.
#4 on Hugging Face · Daily papers, on Oct 30, 09:28 UTC. Trendoday checks this list every 120 minutes.
It was first seen on Oct 30, 2026 and has been on the list on 2 days (14 checks) since.