Salvato in:
| Autori principali: | , , , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2025
|
| Soggetti: | |
| Accesso online: | https://arxiv.org/abs/2512.08524 |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866915663437627392 |
|---|---|
| author | Ahad, Jawad Ibn Rahman, Maisha Biswas, Amrijit Kabir, Muhammad Rafsan Krambroeckers, Robin Momen, Sifat Mohammed, Nabeel Rahman, Shafin |
| author_facet | Ahad, Jawad Ibn Rahman, Maisha Biswas, Amrijit Kabir, Muhammad Rafsan Krambroeckers, Robin Momen, Sifat Mohammed, Nabeel Rahman, Shafin |
| contents | Multimodal language models (MLLMs) require large parameter capacity to align high-dimensional visual features with linguistic representations, making them computationally heavy and difficult to deploy efficiently. We introduce a progressive reparameterization strategy that compresses these models by gradually replacing dense feed-forward network blocks with compact Parameterized Hypercomplex Multiplication (PHM) layers. A residual interpolation schedule, together with lightweight reconstruction and knowledge distillation losses, ensures that the PHM modules inherit the functional behavior of their dense counterparts during training. This transition yields substantial parameter and FLOP reductions while preserving strong multimodal alignment, enabling faster inference without degrading output quality. We evaluate the approach on multiple vision-language models (VLMs). Our method maintains performance comparable to the base models while delivering significant reductions in model size and inference latency. Progressive PHM substitution thus offers an architecture-compatible path toward more efficient multimodal reasoning and complements existing low-bit quantization techniques. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2512_08524 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Beyond Real Weights: Hypercomplex Representations for Stable Quantization Ahad, Jawad Ibn Rahman, Maisha Biswas, Amrijit Kabir, Muhammad Rafsan Krambroeckers, Robin Momen, Sifat Mohammed, Nabeel Rahman, Shafin Computer Vision and Pattern Recognition Computation and Language Multimodal language models (MLLMs) require large parameter capacity to align high-dimensional visual features with linguistic representations, making them computationally heavy and difficult to deploy efficiently. We introduce a progressive reparameterization strategy that compresses these models by gradually replacing dense feed-forward network blocks with compact Parameterized Hypercomplex Multiplication (PHM) layers. A residual interpolation schedule, together with lightweight reconstruction and knowledge distillation losses, ensures that the PHM modules inherit the functional behavior of their dense counterparts during training. This transition yields substantial parameter and FLOP reductions while preserving strong multimodal alignment, enabling faster inference without degrading output quality. We evaluate the approach on multiple vision-language models (VLMs). Our method maintains performance comparable to the base models while delivering significant reductions in model size and inference latency. Progressive PHM substitution thus offers an architecture-compatible path toward more efficient multimodal reasoning and complements existing low-bit quantization techniques. |
| title | Beyond Real Weights: Hypercomplex Representations for Stable Quantization |
| topic | Computer Vision and Pattern Recognition Computation and Language |
| url | https://arxiv.org/abs/2512.08524 |