MoBE: Mixture-of-Basis-Experts for Compressing MoE-based LLMs

• URL: http://arxiv.org/abs/2508.05257v1
• Date: Thu, 07 Aug 2025 10:48:24 GMT
• Title: MoBE: Mixture-of-Basis-Experts for Compressing MoE-based LLMs
• Authors: Xiaodong Chen, Mingming Ha, Zhenzhong Lan, Jing Zhang, Jianguo Li,
• Abstract summary: MoBE achieves notably lower accuracy drops compared to prior works.<n>MoBE can reduce the parameter counts of Qwen3-235B-A22B-2507, DeepSeek-V3-0324 (671B) and Kimi-K2-Instruct (1T) by 24%-30% with only 1%-2% accuracy drop.
• Score: 25.218350580080447
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: The Mixture-of-Experts (MoE) architecture has become a predominant paradigm for scaling large language models (LLMs). Despite offering strong performance and computational efficiency, large MoE-based LLMs like DeepSeek-V3-0324 and Kimi-K2-Instruct present serious challenges due to substantial memory requirements in deployment. While recent works have explored MoE compression to address this issue, existing methods often suffer from considerable accuracy drops (e.g., 7-14% relatively) even at modest compression rates. This paper introduces a novel Mixture-of-Basis-Experts (MoBE) method that achieves model compression while incurring minimal accuracy drops. Specifically, each up/gate matrix in an expert is decomposed via a rank decomposition as W = AB, where matrix A is unique to each expert. The relatively larger matrix B is further re-parameterized as a linear combination of basis matrices {Bi} shared across all experts within a given MoE layer. The factorization is learned by minimizing the reconstruction error relative to the original weight matrices. Experiments demonstrate that MoBE achieves notably lower accuracy drops compared to prior works. For instance, MoBE can reduce the parameter counts of Qwen3-235B-A22B-2507, DeepSeek-V3-0324 (671B) and Kimi-K2-Instruct (1T) by 24%-30% with only 1%-2% accuracy drop (about 2% drops when measured relatively).

Related papers

• MiLo: Efficient Quantized MoE Inference with Mixture of Low-Rank Compensators [17.024303421560578]
  We introduce MiLo, a novel method that augments highly quantized MoEs with a mixture of low-rank compensators.<n>MiLo does not rely on calibration data, allowing it to generalize to different MoE models and datasets without overfitting to a calibration set.<n>Our evaluation shows that MiLo outperforms existing methods on SoTA MoE models across various tasks.
  arXiv  Detail & Related papers  (2025-04-03T14:54:17Z)
• ResMoE: Space-efficient Compression of Mixture of Experts LLMs via Residual Restoration [61.579842548990754]
  Mixture-of-Experts (MoE) Transformer, the backbone of multiple phenomenal language models, leverages sparsity by activating only a fraction of model parameters for each input token.<n>We introduce ResMoE, an innovative MoE approximation framework that utilizes Wasserstein barycenter to extract a common expert (barycenter expert) and approximate the residuals between this barycenter expert and the original ones.
  arXiv  Detail & Related papers  (2025-03-10T03:15:54Z)
• MoE-I$^2$: Compressing Mixture of Experts Models through Inter-Expert Pruning and Intra-Expert Low-Rank Decomposition [32.97035551579975]
  We introduce a two-stage compression method tailored for MoE to reduce the model size and decrease the computational cost. Experiments on Qwen1.5-MoE-A2.7B, DeepSeek-V2-Lite, and Mixtral-8$times$7B demonstrate that our proposed methods can both reduce the model size and enhance inference efficiency.
  arXiv  Detail & Related papers  (2024-11-01T20:37:58Z)
• MoDeGPT: Modular Decomposition for Large Language Model Compression [59.361006801465344]
  This paper introduces textbfModular bfDecomposition (MoDeGPT), a novel structured compression framework.<n>MoDeGPT partitions the Transformer block into modules comprised of matrix pairs and reduces the hidden dimensions.<n>Our experiments show MoDeGPT, without backward propagation, matches or surpasses previous structured compression methods.
  arXiv  Detail & Related papers  (2024-08-19T01:30:14Z)
• From Low Rank Gradient Subspace Stabilization to Low-Rank Weights: Observations, Theories, and Applications [85.17672240603011]
  We study the non-uniform low-rank properties of weight matrices in Large Language Models.<n>We present Weight Low-Rank Projection (WeLore) that unifies weight compression and memory-efficient fine-tuning into one.
  arXiv  Detail & Related papers  (2024-07-15T21:05:20Z)
• QuantMoE-Bench: Examining Post-Training Quantization for Mixture-of-Experts [47.01697456105496]
  Mixture-of-Experts (MoE) is a promising way to scale up the learning capacity of large language models.<n>MoE suffers from significant memory overheads due to the vast parameter size.<n>Post-training quantization offers a powerful approach for model compression.
  arXiv  Detail & Related papers  (2024-06-12T12:44:48Z)
• SliM-LLM: Salience-Driven Mixed-Precision Quantization for Large Language Models [63.118592279833656]
  Post-training quantization (PTQ) is an effective technique for compressing large language models (LLMs)<n>We propose SliM-LLM, a salience-driven mixed-precision quantization framework that allocates bit-widths at the group-wise.<n> Experiments show that SliM-LLM achieves superior performance across various LLMs at low bit-widths.
  arXiv  Detail & Related papers  (2024-05-23T16:21:48Z)
• LORD: Low Rank Decomposition Of Monolingual Code LLMs For One-Shot Compression [16.901290551711476]
  We study the potential to compress Large Language Models (LLMs) for monolingual Code generation via Low Rank Decomposition (LoRD) We then use Low Rank Decomposition (LoRD) to compress StarCoder 16B to 13.2B parameter with no drop and to 12.3B with minimal drop in HumanEval Pass@1 score, in less than 10 minutes on a single A100.
  arXiv  Detail & Related papers  (2023-09-25T10:35:17Z)
• SqueezeLLM: Dense-and-Sparse Quantization [80.32162537942138]
  Main bottleneck for generative inference with LLMs is memory bandwidth, rather than compute, for single batch inference. We introduce SqueezeLLM, a post-training quantization framework that enables lossless compression to ultra-low precisions of up to 3-bit. Our framework incorporates two novel ideas: (i) sensitivity-based non-uniform quantization, which searches for the optimal bit precision assignment based on second-order information; and (ii) the Dense-and-Sparse decomposition that stores outliers and sensitive weight values in an efficient sparse format.
  arXiv  Detail & Related papers  (2023-06-13T08:57:54Z)
• Language model compression with weighted low-rank factorization [73.61874728240568]
  We introduce Fisher information to weigh the importance of parameters affecting the model prediction. We find that our resulting task accuracy is much closer to the original model's performance. Our method can directly compress a task-specific model while achieving better performance than other compact model strategies.
  arXiv  Detail & Related papers  (2022-06-30T21:57:07Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.