Demons in the Detail: On Implementing Load Balancing Loss for Training Specialized Mixture-of-Expert Models (2025.acl-long)
Copied to clipboard
Zihan Qiu, Zeyu Huang, Bo Zheng, Kaiyue Wen, Zekun Wang, Rui Men, Ivan Titov, Dayiheng Liu, Jingren Zhou, Junyang Lin
| Challenge: | Existing Mixture-of-Experts training frameworks use a micro-batch to calculate LBL . micro-batches are restricted to a single sequence, preventing expert specialization . |
| Approach: | They propose to use a global-batch to loosen the load balance constraint for MoEs models . they propose to synchronize fi across micro-batches and then use it to calculate the LBL . |
| Outcome: | The proposed global-batch LBL improves the domain specialization of experts . the micro-battery LBL is almost at the sequence level, and the router is pushed to distribute the token evenly . |
Similar Papers
From Pseudo-Balancing to True Specialization: Memory-Aware Routing for Mixture-of-Experts (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing methods to optimize expert-centered load balancing fail to account for pseudo-balance phenomenon . severe knowledge overlap among experts leads to redundant representations and inefficient parameter utilization . |
| Approach: | They propose a method that prioritizes expert utilization over semantic alignment . they use memory-aware routing to ensure expert load balancing is consistent . |
| Outcome: | Experimental results show that MAR improves expert specialization by 35% and accuracy by 2%-25% . MAR matches baseline performance with only half the experts . |
A Closer Look into Mixture-of-Experts in Large Language Models (2025.findings-naacl)
Copied to clipboard
| Challenge: | Mixture-of-experts (MoE) architectures are gaining increasing attention for their unique properties and remarkable performance. |
| Approach: | They propose a mixture-of-experts architecture that allows for model scaling without sacrificing computational efficiency. |
| Outcome: | The proposed model increases model size without sacrificing computational efficiency . the proposed model is modular and can be used by a broad spectrum of practitioners . |
LLaMA-MoE: Building Mixture-of-Experts from LLaMA with Continual Pre-Training (2024.emnlp-main)
Copied to clipboard
| Challenge: | Mixture-of-Experts (MoE) has gained increasing popularity as a framework for scaling up large language models. |
| Approach: | They investigate how to build Mixture-of-Experts (MoE) models from existing large language models . they use expert construction, Continual pre-training and data sampling strategies . |
| Outcome: | The proposed model outperforms existing models with similar parameters on a wide range of tasks. |
Infinity-MoE: Generalizing Mixture of Experts to Infinite Experts (2026.eacl-short)
Copied to clipboard
| Challenge: | Existing methods to increase the number of experts are -MoE and . |
| Approach: | They propose a mixture of experts that selects a few feed-forward networks per token to increase the number of experts. |
| Outcome: | The proposed model improves on a GPT-2 Small model with 129M active and 186M total parameters by 2.5% over the current model. |
Accelerating Dense LLMs via L0-regularized Mixture-of-Experts (2025.acl-short)
Copied to clipboard
| Challenge: | Existing methods for accelerating large language models (LLMs) suffer from slow and costly inference. |
| Approach: | They propose a lightweight MoE approach using cluster confusion matrix and dynamic batching to accelerate dense LLMs. |
| Outcome: | The proposed method achieves 2.5x speedup over dense models while maintaining competitive performance. |
Bag of Tricks for Sparse Mixture-of-Experts: A Benchmark Across Reasoning, Efficiency, and Safety (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Existing benchmarks focus on isolated aspects of MoE, with conflicting conclusions . a lack of consensus on optimal design choices is limiting to specific aspects of the model. |
| Approach: | They propose to evaluate two popular MoE backbones across four dimensions of design choices . they find token-level routing and z-loss regularization improve reasoning performance . |
| Outcome: | The proposed framework evaluates two popular MoE backbones on over eight metrics. |
Let the Expert Stick to His Last: Expert-Specialized Fine-Tuning for Sparse Architectural Large Language Models (2024.emnlp-main)
Copied to clipboard
| Challenge: | Existing studies on parameter-efficient fine-tuning (PEFT) for dense-architecture LLMs are lacking. |
| Approach: | They propose an expert-specialized fine-tuning method that tunes the experts most relevant to downstream tasks while freezing the other experts. |
| Outcome: | The proposed method matches or surpasses full-parameter fine-tuning. |
Advancing MoE Efficiency: A Collaboration-Constrained Routing (C2R) Strategy for Better Expert Parallelism Design (2025.naacl-long)
Copied to clipboard
| Challenge: | Using Mixture-of-Experts, researchers have found that efficient MoE is difficult to achieve due to two key reasons: imbalanced expert activation and massive communication overhead. |
| Approach: | They propose a collaboration-constrained routing strategy that encourages more specialized expert groups and leverages expert specialization. |
| Outcome: | The proposed approach achieves an average performance improvement of 0.51% and 0.33% on LLaMA-MoE and Qwen-MaE respectively. |
HMoE: Heterogeneous Mixture of Experts for Language Modeling (2025.emnlp-main)
Copied to clipboard
An Wang, Xingwu Sun, Ruobing Xie, Shuaipeng Li, Jiaqi Zhu, Zhen Yang, Pinxue Zhao, Weidong Han, Zhanhui Kang, Di Wang, Naoaki Okazaki, Cheng-zhong Xu
| Challenge: | Mixture of Experts (MoE) models use homogeneous experts with diverse capacities, resulting in a lack of expert specialization and parameter utilization. |
| Approach: | They propose a framework where experts differ in size and possess diverse capacities . they propose HMoE to encourage frequent activation of smaller experts . |
| Outcome: | The proposed framework outperforms homogeneous homogenous MoE models on evaluation benchmarks and achieves lower loss rate with fewer activated parameters. |
Alloc-MoE: Budget-Aware Expert Activation Allocation for Efficient Mixture-of-Experts Inference (2026.acl-long)
Copied to clipboard
| Challenge: | Existing approaches that reduce expert activations lead to severe model performance degradation. |
| Approach: | They propose a framework that optimizes budget allocation coordinately at layer and token levels to minimize model performance degradation. |
| Outcome: | The proposed framework achieves 1.15 prefill and 1.34 decode speedups on DeepSeek-V2-Lite at half of the original budget. |