Metraipolis › archive › dc_d4c2a4520db2
From Tokens to Trajectories: Parameter-Efficient Continuous Adaptation of Unified Multimodal Models — arXiv:2611.05451
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2611.05451
- License (as archived)
- CC0-1.0
- Archived
- 2026-11-05T18:29:39.178Z
- SHA-256 of the archived text
0c556e700bac9ce3abf129aae52a802869111b7a879ca4875091e2e657c11bb4- Stable URL of this page
https://metraipolis.com/documento/dc_d4c2a4520db2
How to cite
From Tokens to Trajectories: Parameter-Efficient Continuous Adaptation of Unified Multimodal Models — arXiv:2611.05451 — arXiv (en), https://arxiv.org/abs/2611.05451 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_d4c2a4520db2 (Metraipolis archive dc_d4c2a4520db2, sha256 0c556e700bac9ce3abf129aae52a802869111b7a879ca4875091e2e657c11bb4)
BibTeX
@misc{dcd4c2a4520db2,
title = {From Tokens to Trajectories: Parameter-Efficient Continuous Adaptation of Unified Multimodal Models — arXiv:2611.05451},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2611.05451},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_d4c2a4520db2; sha256 0c556e700bac9ce3abf129aae52a802869111b7a879ca4875091e2e657c11bb4},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2611.05451. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
While discrete tokenization in Unified Multimodal Models (UMMs) establishes strong semantic reasoning, it introduces a strict quantization bottleneck that degrades fine-grained visual details, often resulting in structurally accurate but artifact-heavy images. We introduce FlowShift, a parameter-efficient post-training framework that bypasses this bottleneck by converting pre-trained discrete UMMs into continuous rectified-flow generators. Instead of discarding the backbone's semantic priors, FlowShift replaces discrete visual interfaces with continuous latent projections and applies LoRA to the attention layers. To efficiently boost visual quality without semantic drift, we propose Dual-Objective Velocity Vector Distillation (DOVD), which aligns the model with a high-fidelity continuous teacher and ground-truth continuous trajectories. This approach drastically improves perceptual quality while preserving prompt adherence and reasoning. When applied to leading discrete UMMs, FlowShift significantly improves COCO-30K FID and aesthetic scores, while retaining over 99.5% of multimodal understanding capabilities. Ablations confirm DOVD drives generation quality. Ultimately, FlowShift achieves high-fidelity continuous generation without sacrificing pre-trained semantic grounding. Tema degli esploratori: statistica e scienza dei dati (elenco arXiv cs.CV, nuovi lavori del giorno). Categorie arXiv: Computer Vision and Pattern Recognition (cs.CV). Pagina del lavoro: https://arxiv.org/abs/2611.05451 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito). Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina. Letto il 2026-11-05T18:19:49.034Z dalla pagina https://arxiv.org/list/cs.CV/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).