Highly Efficient and Effective LLMs with Multi-Boolean Architectures

arXiv:2505.22811v5 Announce Type: replace-cross Abstract: Weight binarization has emerged as a promising strategy to reduce the complexity of large language models (LLMs). Existing approaches fall into post-training binarization, which is simple but causes severe performance loss, and training-aware methods, which depend on full-precision latent weights, adding complexity and limiting efficiency. We propose a novel framework that represents LLMs with multi-kernel Boolean parameters and, for the first time, enables direct finetuning LMMs in the Boolean domain, eliminating the need for latent weights. This enhances representational capacity and dramatically reduces complexity during both finetuning and inference. Extensive experiments across diverse LLMs show our method outperforms recent ultra low-bit quantization and binarization techniques.

Related

• From 2:4 to 8:16 sparsity patterns in LLMs for Outliers and Weights with Variance Correction
• OASIS: Online Activation Subspace Learning for Memory-Efficient Training
• YOCO++: Enhancing YOCO with KV Residual Connections for Efficient LLM Inference
• QSLM: A Performance- and Memory-aware Quantization Framework with Tiered Search Strategy for Spike-driven Language Models

Source: arXiv cs.LG | 2026-04-22