DEMix Layers: Disentangling Domains for Modular Language Modeling (2022.naacl-main)
Copied to clipboard
| Challenge: | Extensive experiments with autoregressive transformer LMs show that DEMix layers reduce test-time perplexity and increase training efficiency. |
| Approach: | They introduce a new domain expert mixture layer that enables conditioning a language model on the domain of the input text. |
| Outcome: | Experiments with 1.3B LMs show that DEMix layers reduce test-time perplexity, increase training efficiency, and enable rapid adaptation. |
Similar Papers
Injecting Domain-Specific Knowledge into Large Language Models: A Comprehensive Survey (2025.findings-emnlp)
Copied to clipboard
| Challenge: | specialized LLMs are often limited in domain-specific applications that require specialized knowledge. |
| Approach: | They provide a comprehensive overview of four key methods to enhance large language models by integrating domain-specific knowledge. |
| Outcome: | The proposed methods are categorized into four key approaches: dynamic knowledge injection, static knowledge embedding, modular adapters, and prompt optimization. |
A synthetic data approach for domain generalization of NLI models (2024.acl-long)
Copied to clipboard
| Challenge: | Natural Language Inference (NLI) datasets are important benchmark tasks for LLMs . however, their realistic performance on out-of-distribution/domain data is less well-understood . a T5-small model trained with our data improves around 7% on average compared to the best alternative dataset . |
| Approach: | They propose a new approach for generating NLI data in diverse domains and lengths . they show that models trained on this data have the best generalization to completely new downstream test settings . |
| Outcome: | The proposed model can be trained on datasets with high-quality examples with meaningful premises and high accuracy. |
ModularMoE: Fast LLM Customization with Parameter-Sharing Mixture-of-Experts for Low-Resource Settings (2026.findings-acl)
Copied to clipboard
Jiaxing Liu, Qi Qi, Haifeng Sun, Dunjun Li, Zirui Zhuang, Bo He, Xiang Yang, Cong Liu, Jianxin Liao, Jingyu Wang
| Challenge: | Large Language Models impose significant computational and storage burdens on personal devices . existing customization approaches incur excessive computational costs or lead to suboptimal performance . |
| Approach: | They propose a training framework that converts pre-trained LLMs into parameter-sharing MoE models for lightweight deployment. |
| Outcome: | The proposed training framework outperforms state-of-the-art training frameworks at the same sparsity level while delivering up to 2.71 inference speedup. |
Unlocking Emergent Modularity in Large Language Models (2024.naacl-long)
Copied to clipboard
| Challenge: | Existing MNNs are explicit, with predefined modular architectures and individual modules expected to implement distinct functions. |
| Approach: | They propose to unlock emergent modularity in language models by fine-tuning them as Mixture-of-Experts (MoEs) EMoE is robust to various configurations and can scale up to Large Language Models . |
| Outcome: | The proposed models can be fine-tuned as Mixture-of-Expert (MoE) counterparts without introducing any extra parameters. |
Do Domain-specific Experts exist in MoE-based LLMs? (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing studies on domain-specific experts in Large Language Models (LLMs) are still lacking. |
| Approach: | They propose a training-free framework that introduces zero additional inference cost and outperforms well-trained MoE-based LLMs. |
| Outcome: | The proposed framework outperforms well-trained MoE-based LLMs and strong baselines across target and non-target domains. |
Less, but Better: Efficient Multilingual Expansion for LLMs via Layer-wise Mixture-of-Experts (2025.acl-long)
Copied to clipboard
| Challenge: | Existing large language models (LLMs) have remarkable ability in high-resource languages, but their performance in multilingual scenarios is still limited. |
| Approach: | They propose a layer-wise expert allocation algorithm to determine the appropriate number of new experts for each layer. |
| Outcome: | The proposed method outperforms the previous state-of-the-art baseline with 60% fewer experts in the single-expansion setting and 33.3% fewer in the lifelong-expanding setting. |
Self-Specialization: Uncovering Latent Expertise within Large Language Models (2024.findings-acl)
Copied to clipboard
Junmo Kang, Hongyin Luo, Yada Zhu, Jacob Hansen, James Glass, David Cox, Alan Ritter, Rogerio Feris, Leonid Karlinsky
| Challenge: | Recent studies have demonstrated the effectiveness of self-alignment in which a large language model is aligned to follow general instructions using instructional data generated from the model itself. |
| Approach: | They propose to use human-written seeds to align large language models to follow general instructions to achieve cross-task generalization. |
| Outcome: | The proposed model outperforms base models and models that are generally instruction-tuned or have been adapted to the target domain by a large margin. |
Split-Merge: Scalable and Memory-Efficient Merging of Expert LLMs (2025.emnlp-main)
Copied to clipboard
| Challenge: | a zero-shot merging framework for large language models consolidates specialized domain experts into a single model without any further training. |
| Approach: | They propose a zero-shot merging framework that consolidates specialized domain experts into a single model without further training. |
| Outcome: | Experiments on code generation, mathematical reasoning, medical question answering, and instruction-following benchmarks confirm the versatility and effectiveness of the proposed framework. |
Specialization through Collaboration: Understanding Expert Interaction in Mixture-of-Expert Large Language Models (2026.eacl-long)
Copied to clipboard
| Challenge: | Mixture-of-Experts (MoE) based large language models are popular for multitasking . however, whether each expert can specialize to a task remains unclear . |
| Approach: | They propose to use a dictionary learning approach to analyze expert collaboration mechanisms in MoE LLMs. |
| Outcome: | The proposed model outperforms existing methods by 2.5% while enabling 50% expert reduction. |
MergeME: Model Merging Techniques for Homogeneous and Heterogeneous MoEs (2025.naacl-long)
Copied to clipboard
Yuhang Zhou, Giannis Karamanolakis, Victor Soto, Anna Rumshisky, Mayank Kulkarni, Furong Huang, Wei Ai, Jianhua Lu
| Challenge: | State-of-the-art methods for merging expert models with different architectures do not address parameter interference and require extensive fine-tuning to restore performance. |
| Approach: | They propose a method for merging experts with different architectures into a unified Mixture-of-Experts model with a goal of enhancing performance in each domain while retaining effectiveness on general tasks. |
| Outcome: | Experiments across multiple domains show that the proposed methods reduce fine-tuning costs and improve performance over state-of-the-art methods. |