This work proposes MoEGen, an adaptation framework that shifts MoE-based PEFT from expert selection to expert-conditioned parameter generation, and decouples expert capacity from adapter storage while enabling instance-conditioned adaptation.
Abstract
Parameter-efficient fine-tuning (PEFT) enables efficient adaptation of large language models, but existing MoE-based PEFT methods typically improve capacity by storing multiple full LoRA experts, causing adapter storage to grow linearly with the number of experts and restricting adaptation to a fixed expert pool. We ask whether MoE-based PEFT can produce instance-specific adaptations without explicitly storing a separate LoRA module for each expert. To address this gap, we propose MoEGen, an adaptation framework that shifts MoE-based PEFT from expert selection to expert-conditioned parameter generation. Instead of storing each expert as a full LoRA adapter, MoEGen represents each expert as a small learnable vector, termed an expert code. It routes each input over these vectors and uses their weighted combination to condition a lightweight hypernetwork that generates input-specific low-rank updates. This design decouples expert capacity from adapter storage while enabling instance-conditioned adaptation. Experiments on eight commonsense reasoning benchmarks show consistent improvements over strong static and MoE-based PEFT baselines across three backbones. MoEGen also performs strongly in joint medical and legal-domain adaptation.
This paper makes the first attempt to fine-tune MoE models with MoE-style low-rank adaptation via a dual-channel Routing-Conditioned Projection module, which reuses base router activations to inform LoRA routing and introduces a single global LoRA expert pool shared across all layers.
Qingyu Yang, Haonan He, Minglei Li et al.· arXiv.org· 0 citations
An algorithm framework named DR-EFT (Domain-Representative Experts for Fine-Tuning), which explores and loads the domain-representative experts for subsequent retraining and reincorporation and demonstrates robustness through validations on popular MoE LLMs, including Qwen, DeepSeek, and Ernie.
Zhaomeng Cheng, Zhong Ji, Yan Zhang et al.· Neural Networks· 0 citations
Mixture-of-Experts (MoE) language models route each token through a small subset of experts, making routing patterns useful for identifying task-relevant experts during downstream adaptation. Yet current approaches have two limitations: task experts are typically identified from aggregate routing statistics that reflect usage rather than association with successful task completion, and task-expert activations remain underexplored as signals for supervision allocation. We introduce Task-Expert-Aware Supervision (TEXAS), which combines correctness-conditioned task expert discovery with token-level supervision allocation. TEXAS compares expert activations on instances that the base model solves successfully and those it fails to solve, and retains experts more strongly activated on successful instances. During fine-tuning, it upweights answer tokens in failed instances when they activate these experts. TEXAS therefore leverages existing routing behavior without restricting adaptation to a fixed expert subset or imposing an explicit target routing distribution. Across three MoE models and six benchmarks, TEXAS achieves the best or tied-best performance in 17 of 18 settings and improves over the strongest baseline by 1.3--1.5 points on average. Ablations and further analyses validate both the discovered experts and the resulting supervision strategy.
Guanzhi Deng, Haibo Wang, Kuan Wu et al.· 0 citations
Communication-efficient MoE models (CE-MoE), in which a heterogeneous layer pattern that decouples token-mixing and channel-mixing depth is adopted, consistently reduce training cost while matching validation loss and downstream benchmarks with full-MoE baselines.
MetaNet is proposed, a support-set controller that predicts, for each layer, an expert-retention threshold and a bounded routing bias and provides a tunable accuracy-expert-activation trade-off on DeepSeek-MoE-16B-Chat.
Rong-Feng Wang, Shitao Weng, Zhiquan Wang et al.· 0 citations
Mixture-of-Experts (MoE) architectures provide an efficient paradigm for scaling large language models (LLMs), yet fixed top-k routing activates the same number of expert slots for every token, causing substantial redundant computation. Existing expert-skipping methods often rely on router confidence, calibration data, or additional training, and therefore cannot reliably estimate the actual contribution of routed experts. To this end, we propose ACE, a training-free, calibration-free, and checkpoint-preserving framework for token-adaptive expert skipping in MoE-based LLMs. ACE contains two complementary components: 1) Global Spectral Proxy (GSP), which estimates global transformation capacity from the coupled gate, up, and down projections together with RMSNorm scaling; and 2) Router-Conditioned Refinement (RCR), which constructs expert-specific direction prototypes from centered router weights and evaluates expert responses along routing-preferred directions. During inference, ACE combines both estimates with runtime router gates and skips an expert slot only when both views identify it as low-contribution, while always retaining the top-1 expert. All expert statistics are computed offline, leaving only table lookups and lightweight scalar operations online. Extensive experiments across three MoE-based LLMs and eight benchmarks demonstrate that ACE consistently outperforms existing static and dynamic baselines, with increasingly pronounced advantages under aggressive expert skipping. For instance, at a 50% skipping ratio on Qwen3.6-35B-A3B, ACE reduces WikiText-2 perplexity by 7.96% and improves average downstream accuracy by 4.15 percentage points over the strongest competing method.
Zukang Xu, Zhixiong Zhao, Xing Hu et al.· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
