Beyond Real Weights: Hypercomplex Representations for Stable Quantization

TL;DR

This paper presents a progressive reparameterization method using hypercomplex layers to compress multimodal language models, reducing parameters and computation while maintaining performance for efficient multimodal reasoning.

Contribution

Introduces a novel progressive replacement of dense layers with PHM layers, enabling efficient model compression without performance loss.

Findings

Significant parameter and FLOP reduction achieved.

Maintains model performance comparable to original models.

Enables faster inference for multimodal models.

Abstract

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.