Metraipolis › archive › dc_2a142fed5ad5

NeuPAT: Neuron-aware Plasticity Allocation Tuning for Language-Preserving MLLMs — arXiv:2608.08107

Source
arXiv (en)
Original
https://arxiv.org/abs/2608.08107
License (as archived)
CC0-1.0
Archived
2026-10-19T08:49:53.900Z
SHA-256 of the archived text
7793df8498cfcf11893c8730940ae801d800ed8603da3923deed0c97de84be60
Stable URL of this page
https://metraipolis.com/documento/dc_2a142fed5ad5

How to cite

NeuPAT: Neuron-aware Plasticity Allocation Tuning for Language-Preserving MLLMs — arXiv:2608.08107 — arXiv (en), https://arxiv.org/abs/2608.08107 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_2a142fed5ad5 (Metraipolis archive dc_2a142fed5ad5, sha256 7793df8498cfcf11893c8730940ae801d800ed8603da3923deed0c97de84be60)
BibTeX
@misc{dc2a142fed5ad5,
  title = {NeuPAT: Neuron-aware Plasticity Allocation Tuning for Language-Preserving MLLMs — arXiv:2608.08107},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2608.08107},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_2a142fed5ad5; sha256 7793df8498cfcf11893c8730940ae801d800ed8603da3923deed0c97de84be60},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.08107. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Multimodal expansion of large language models (LLMs) enables new perceptual capabilities but often compromises the language intelligence acquired during pretraining. In this work, we investigate this phenomenon from the perspective of internal adaptation dynamics and discover that neurons in pretrained LLMs exhibit heterogeneous plasticity during multimodal learning: some neurons are critical for preserving language capabilities, while others are more adaptive to multimodal knowledge. Based on this insight, we propose NeuPAT (Neuron-aware Plasticity Allocation Tuning), a lightweight and architecture-agnostic framework that allocates neuron-wise update constraints during multimodal instruction tuning. NeuPAT uses a small-scale probing stage to estimate neuron adaptation patterns and selectively protects language-sensitive neurons while promoting multimodal adaptation through more plastic neurons. Experiments across diverse LLM families demonstrate that NeuPAT recovers 94.5\% of the language capability degradation caused by vanilla tuning on 11 language benchmarks while maintaining comparable multimodal performance, providing an effective approach for capability-preserving multimodal expansion.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-08-11 in cs.CL).
Categorie arXiv: Computation and Language (cs.CL) ; Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2608.08107 — 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 2026-10-19T08:49:41.848Z dalla pagina https://arxiv.org/catchup/cs.CL/2026-08-11?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).