REAM: Merging Improves Pruning of Experts in LLMs

• URL: http://arxiv.org/abs/2604.04356v1
• Date: Mon, 06 Apr 2026 02:08:09 GMT
• Title: REAM: Merging Improves Pruning of Experts in LLMs
• Abstract summary: Traditional approaches to reduce memory requirements include weight pruning and quantization.<n>We propose a novel method, Router-weighted Expert Activation Merging (REAM)<n>Instead of removing experts, REAM groups them and merges their weights, better preserving original performance.
• Score: 6.52958560748865
• License: http://creativecommons.org/licenses/by-nc-sa/4.0/
• Abstract: Mixture-of-Experts (MoE) large language models (LLMs) are among the top-performing architectures. The largest models, often with hundreds of billions of parameters, pose significant memory challenges for deployment. Traditional approaches to reduce memory requirements include weight pruning and quantization. Motivated by the Router-weighted Expert Activation Pruning (REAP) that prunes experts, we propose a novel method, Router-weighted Expert Activation Merging (REAM). Instead of removing experts, REAM groups them and merges their weights, better preserving original performance. We evaluate REAM against REAP and other baselines across multiple MoE LLMs on diverse multiple-choice (MC) question answering and generative (GEN) benchmarks. Our results reveal a trade-off between MC and GEN performance that depends on the mix of calibration data. By controlling the mix of general, math and coding data, we examine the Pareto frontier of this trade-off and show that REAM often outperforms the baselines and in many cases is comparable to the original uncompressed models.

Related papers

• Share First, Route What Remains: A Unified Framework for Token-Adaptive MoE Computation [49.8459545291965]
  Mixture-of-experts (MoE) models have recently moved beyond routing a fixed number of complete experts.<n>We study this dependency by decomposing sparsely upcycled feed-forward experts into key-value channels.<n>We instantiate it in UniF-MoE, a unified framework for token-adaptive MoE computation.
  arXiv  Detail & Related papers  (2026-08-11T02:40:58Z)
• Tying the Loop -- Tied Expert Layers in Mixture-of-Experts Language Models [50.34585122347149]
  We introduce Expert Tying, an architectural modification that shares expert parameters across consecutive transformer layers.<n>We evaluate this approach across common, state-of-the-art architectures, including OLMoE, Qwen3, and DeepSeek-style MoEs.
  arXiv  Detail & Related papers  (2026-06-15T15:08:09Z)
• Think Then Embed: Generative Context Improves Multimodal Embedding [51.76690812535934]
  We propose a Think-Then-Embed (TTE) framework for Universal Multimodal Embeddings (UME), composed of a reasoner and an embedder.<n>By leveraging a powerful MLLM reasoner, we achieve state-of-the-art performance on the MMEB-V2 benchmark, surpassing proprietary models trained on massive in-house datasets.
  arXiv  Detail & Related papers  (2025-10-06T16:53:56Z)
• DreamPRM: Domain-Reweighted Process Reward Model for Multimodal Reasoning [32.943971698086735]
  We introduce DreamPRM, a domain-reweighted training framework for multimodal PRMs.<n>In the lower-level optimization, DreamPRM performs fine-tuning on multiple datasets with domain weights.<n>In the upper-level optimization, the PRM is evaluated on a separate meta-learning dataset.
  arXiv  Detail & Related papers  (2025-05-26T17:20:17Z)
• Mediator: Memory-efficient LLM Merging with Less Parameter Conflicts and Uncertainty Based Routing [19.681042016834187]
  Model merging aggregates Large Language Models finetuned on different tasks into a stronger one.<n>In this work, we observe that different layers exhibit varying levels of parameter conflicts.<n>We build on this insight and use a novel task-level expert routing for layers with significant conflicts.
  arXiv  Detail & Related papers  (2025-02-06T11:26:30Z)
• Read-ME: Refactorizing LLMs as Router-Decoupled Mixture of Experts with System Co-Design [59.00758127310582]
  We propose a novel framework Read-ME that transforms pre-trained dense LLMs into smaller MoE models. Our approach employs activation sparsity to extract experts. Read-ME outperforms other popular open-source dense models of similar scales.
  arXiv  Detail & Related papers  (2024-10-24T19:48:51Z)
• Your Mixture-of-Experts LLM Is Secretly an Embedding Model For Free [21.59456761618456]
  Large language models (LLMs) excel on generation tasks, their decoder-only architecture often limits their potential as embedding models if no further representation finetuning is applied. Our study shows that the expert routers in MoE LLMs can serve as an off-the-shelf embedding model with promising performance on a diverse class of embedding-focused tasks.
  arXiv  Detail & Related papers  (2024-10-14T17:59:44Z)
• BAM! Just Like That: Simple and Efficient Parameter Upcycling for Mixture of Experts [41.83123857437985]
  Training MoEs from scratch in a large-scale regime is prohibitively expensive. We propose BAM (Branch-Attend-Mix), a simple yet effective method that addresses this shortcoming. Our experiments on seed models ranging from 590 million to 2 billion parameters demonstrate that BAM surpasses baselines in both perplexity and downstream task performance.
  arXiv  Detail & Related papers  (2024-08-15T17:19:12Z)
• Diversifying the Expert Knowledge for Task-Agnostic Pruning in Sparse Mixture-of-Experts [75.85448576746373]
  We propose a method of grouping and pruning similar experts to improve the model's parameter efficiency.<n>We validate the effectiveness of our method by pruning three state-of-the-art MoE architectures.<n>The evaluation shows that our method outperforms other model pruning methods on a range of natural language tasks.
  arXiv  Detail & Related papers  (2024-07-12T17:25:02Z)
• ME-Switch: A Memory-Efficient Expert Switching Framework for Large Language Models [43.29533894162248]
  LLM development involves pre-training a foundation model on massive data, followed by fine-tuning on task-specific data to create specialized experts. Previous approaches decompose the expert weights as the pre-trained weights plus delta weights, followed by quantizing the delta weights to reduce the model size. We introduce ME-Switch, a memory-efficient expert switching framework tailored for serving multiple LLMs.
  arXiv  Detail & Related papers  (2024-06-13T12:27:55Z)
• Omni-SMoLA: Boosting Generalist Multimodal Models with Soft Mixture of Low-rank Experts [74.40198929049959]
  Large multi-modal models (LMMs) exhibit remarkable performance across numerous tasks. generalist LMMs often suffer from performance degradation when tuned over a large collection of tasks. We propose Omni-SMoLA, an architecture that uses the Soft MoE approach to mix many multimodal low rank experts.
  arXiv  Detail & Related papers  (2023-12-01T23:04:27Z)

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.