DualToken: Towards Unifying Visual Understanding and Generation with Dual Visual Vocabularies
arxiv.org/abs/2503.14324

state-of-the-art performance in both reconstruction and semantic tasks while demonstrating remarkable effectiveness in downstream MLLM understanding and generation tasks. Notably, we also show that DualToken, as a unified tokenizer, surpasses the [6/7 of https://arxiv.org/abs/2503.14324v1]

semantic and perceptual information, DualToken disentangles them by introducing separate codebooks for high and low-level features, effectively transforming their inherent conflict into a synergistic relationship. As a result, DualToken achieves [5/7 of https://arxiv.org/abs/2503.14324v1]

learning aligns well with language but struggles to decode back into the pixel space for generation tasks. To bridge this gap, we propose DualToken, a method that unifies representations for both understanding and generation within a single tokenizer. [3/7 of https://arxiv.org/abs/2503.14324v1]

Song, Wang, Song, Li, Sun, Chen, Zhou, Xu, Wang, Yu: DualToken: Towards Unifying Visual Understanding and Generation with Dual Visual Vocabularies https://arxiv.org/abs/2503.14324 https://arxiv.org/pdf/2503.14324 https://arxiv.org/html/2503.14324

Wei Song, Yuran Wang, Zijia Song, Yadong Li, Haoze Sun, Weipeng Chen, Zenan Zhou, Jianhua Xu, Jiaqi Wang, Kaicheng Yu
DualToken: Towards Unifying Visual Understanding and Generation with Dual Visual Vocabularies
https://arxiv.org/abs/2503.14324

$VTHO trading live su #KuCoin!

🚀VTHO/USDT: trade.kucoin.com/VTHO-USDT?utm_source=community_partner_info

Scopri di più su VeThor Token nella #KuCoinCryptoGem card

#VECHAIN #DUALTOKEN