Metraipolis › archive › dc_50c8dfe33ae9

Where a New Concept Must Enter: Entry Point Gates Cross-Task Usability in Unified Multimodal Models — arXiv:2608.17564

Source
arXiv (en)
Original
https://arxiv.org/abs/2608.17564
License (as archived)
CC0-1.0
Archived
2026-10-18T14:33:34.022Z
SHA-256 of the archived text
66df6a5ea305b10cbca92364420582aef74081424621e995a5eb7c85d1c64ecf
Stable URL of this page
https://metraipolis.com/documento/dc_50c8dfe33ae9

How to cite

Where a New Concept Must Enter: Entry Point Gates Cross-Task Usability in Unified Multimodal Models — arXiv:2608.17564 — arXiv (en), https://arxiv.org/abs/2608.17564 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_50c8dfe33ae9 (Metraipolis archive dc_50c8dfe33ae9, sha256 66df6a5ea305b10cbca92364420582aef74081424621e995a5eb7c85d1c64ecf)
BibTeX
@misc{dc50c8dfe33ae9,
  title = {Where a New Concept Must Enter: Entry Point Gates Cross-Task Usability in Unified Multimodal Models — arXiv:2608.17564},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2608.17564},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_50c8dfe33ae9; sha256 66df6a5ea305b10cbca92364420582aef74081424621e995a5eb7c85d1c64ecf},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.17564. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Unified multimodal models (UMMs) are motivated by the hope that understanding and generation reinforce each other but controlled ablations repeatedly find that adding a generation objective leaves understanding flat. Joint-training studies cannot settle the disagreement: with overlapping supervision, a gain cannot be attributed to the architecture rather than the data. To further investigate the relationship between the two directions in UMMs, we separate them by construction. A novel visual entity, a rendered 3D asset paired with a pseudo-word screened for absence from the frozen model's behavior, is bound through exactly one task direction, and the untrained direction is then measured. We find that the channel is real in both directions, but the directions differ in kind: generation training installs a name the model can only match among candidates; understanding training installs one it can also produce. What governs cross-task usability is where the binding enters the shared computation. An alignment probe predicts export across 36 configurations (Spearman $\rho = +0.68$). That objective's alignment term, maximized in closed form over activations with every weight frozen, makes a concept drawable when injected at layer 7 of 28 and is indistinguishable from the base model from layer 14 on, while the weight-based version of the same edit peaks at layers 10-14. In an observational series of four models, this window appears only where the understanding pathway is a semantic vision encoder, suggesting that unified weights are not enough: the two directions must share a semantic format at the entry point. Exploiting the rule, a mid-stack alignment objective acquires the concept for a $0.1\%$ relative loss of the model's general text-to-image ability, against $41\%$ for the standard generative route. Our code is at this https URL .

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-08-19 in cs.AI).
Categorie arXiv: Computer Vision and Pattern Recognition (cs.CV) ; Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2608.17564 — 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 2026-10-18T14:27:54.436Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-08-19?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).