Mixture-of-Experts Models

Mixture of Experts (MoE) models enable parameter-efficient scaling through sparse expert activations, yet optimizing their inference and memory costs remains challenging due to limited understanding of their specialization behavior. We present a systematic analysis of expert specialization in MoEs through two complementary approaches: domain-specific routing patterns and an early decoding framework that tracks expert contributions to output representations. Our analysis of the DeepSeekMoE model reveals that despite having 64 routed experts with 6 active for each layer's computation, the model predominantly relies on a few specialized experts, with the top-weighted expert's output closely approximating the full ensemble prediction. We quantitatively validate these findings through a systematic analysis of the token routing distribution, demonstrating that very few experts handle over 50% of routing decisions across different specialized domains. Hidden state similarity between single and ensemble experts for every layer is extremely high, with some layers having cosine similarity as high as 0.95 and perplexity increasing by only 5% when using a single expert across all three domains. Our results indicate that Mixture of Experts models exhibit concentrated expertise highlighting potential opportunities for inference optimization through targeted expert pruning while maintaining model performance and opening avenues towards studying localization of learned knowledge in these models.

Mixture of Experts (MoE) achieve parameter-efficient scaling through sparse expert routing, yet their internal representations remain poorly understood compared to dense models. We present a systematic comparison of MoE and dense model internals using crosscoders, a variant of sparse autoencoders, that jointly models multiple activation spaces. We train 5-layer dense and MoEs (equal active parameters) on 1B tokens across code, scientific text, and english stories. Using BatchTopK crosscoders with explicitly designated shared features, we achieve $\sim 87\%$ fractional variance explained and uncover concrete differences in feature organization. The MoE learns significantly fewer unique features compared to the dense model. MoE-specific features also exhibit higher activation density than shared features, whereas dense-specific features show lower density. Our analysis reveals that MoEs develop more specialized, focused representations while dense models distribute information across broader, more general-purpose features.

Mixture-of-Experts (MoE) decouples model capacity from per-token computation, yet their scalability remains limited by the physical dimensions of depth and width. To overcome this, we propose Mixture of Universal Experts (MOUE),a MoE generalization introducing a novel scaling dimension: Virtual Width. In general, MoUE aims to reuse a universal layer-agnostic expert pool across layers, converting depth into virtual width under a fixed per-token activation budget. However, two challenges remain: a routing path explosion from recursive expert reuse, and a mismatch between the exposure induced by reuse and the conventional load-balancing objectives. We address these with three core components: a Staggered Rotational Topology for structured expert sharing, a Universal Expert Load Balance for depth-aware exposure correction, and a Universal Router with lightweight trajectory state for coherent multi-step routing. Empirically, MoUE consistently outperforms matched MoE baselines by up to 1.3% across scaling regimes, enables progressive conversion of existing MoE checkpoints with up to 4.2% gains, and reveals a new scaling dimension for MoE architectures.

We introduce Timer-S1, a strong Mixture-of-Experts (MoE) time series foundation model with 8.3B total parameters, 0.75B activated parameters for each token, and a context length of 11.5K. To overcome the scalability bottleneck in existing pre-trained time series foundation models, we perform Serial Scaling in three dimensions: model architecture, dataset, and training pipeline. Timer-S1 integrates sparse TimeMoE blocks and generic TimeSTP blocks for Serial-Token Prediction (STP), a generic training objective that adheres to the serial nature of forecasting. The proposed paradigm introduces serial computations to improve long-term predictions while avoiding costly rolling-style inference and pronounced error accumulation in the standard next-token prediction. Pursuing a high-quality and unbiased training dataset, we curate TimeBench, a corpus with one trillion time points, and apply meticulous data augmentation to mitigate predictive bias. We further pioneer a post-training stage, including continued pre-training and long-context extension, to enhance short-term and long-context performance. Evaluated on the large-scale GIFT-Eval leaderboard, Timer-S1 achieves state-of-the-art forecasting performance, attaining the best MASE and CRPS scores as a pre-trained model. Timer-S1 is released to facilitate further research.

Electrocardiography (ECG) analysis is crucial for cardiac diagnosis, yet existing foundation models often fail to capture the periodicity and diverse features required for varied clinical tasks. We propose ECG-MoE, a hybrid architecture that integrates multi-model temporal features with a cardiac period-aware expert module. Our approach uses a dual-path Mixture-of-Experts to separately model beat-level morphology and rhythm, combined with a hierarchical fusion network using LoRA for efficient inference. Evaluated on five public clinical tasks, ECG-MoE achieves state-of-the-art performance with 40% faster inference than multi-task baselines.

Multimodal Transformers often produce predictions without clarifying how different modalities jointly support a decision. Most existing multimodal explainable AI (MXAI) methods extend unimodal saliency to multimodal backbones, highlighting important tokens or patches within each modality, but they rarely pinpoint which cross-modal feature pairs provide complementary evidence (synergy) or serve as reliable backups (redundancy). We present Feature-level I2MoE (FL-I2MoE), a structured Mixture-of-Experts layer that operates directly on token/patch sequences from frozen pretrained encoders and explicitly separates unique, synergistic, and redundant evidence at the feature level. We further develop an expert-wise explanation pipeline that combines attribution with top-K% masking to assess faithfulness, and we introduce Monte Carlo interaction probes to quantify pairwise behavior: the Shapley Interaction Index (SII) to score synergistic pairs and a redundancy-gap score to capture substitutable (redundant) pairs. Across three benchmarks (MMIMDb, ENRICO, and MMHS150K), FL-I2MoE yields more interactionspecific and concentrated importance patterns than a dense Transformer with the same encoders. Finally, pair-level masking shows that removing pairs ranked by SII or redundancy-gap degrades performance more than masking randomly chosen pairs under the same budget, supporting that the identified interactions are causally relevant. Code is available at https://github.com/dut0817/FL-I2MoE.

Pathology report generation remains a relatively under-explored downstream task, primarily due to the gigapixel scale and complex morphological heterogeneity of Whole Slide Images (WSIs). Existing pathology report generation frameworks typically employ transformer architectures, relying on a homogeneous decoder architecture and static knowledge retrieval integration. Such architectures limit generative specialization and may introduce noisy external guidance during the report generation process. To address these limitations, we propose RANGER, a sparsely-gated Mixture-of-Experts (MoE) framework with adaptive retrieval re-ranking for pathology report generation. Specifically, we integrate a sparsely gated MoE into the decoder, along with noisy top-$k$ routing and load-balancing regularization, to enable dynamic expert specialization across various diagnostic patterns. Additionally, we introduce an adaptive retrieval re-ranking module that selectively refines retrieved memory from a knowledge base before integration, reducing noise and improving semantic alignment based on visual feature representations. We perform extensive experiments on the PathText-BRCA dataset and demonstrate consistent improvements over existing approaches across standard natural language generation metrics. Our full RANGER model achieves optimal performance on PathText dataset, reaching BLEU-1 to BLEU-4 scores of 0.4598, 0.3044, 0.2036, and 0.1435, respectively, with METEOR of 0.1883, and ROUGE-L of 0.3038, validating the effectiveness of dynamic expert routing and adaptive knowledge refinement for semantically grounded pathology report generation.

For effective deployment in real-world environments, humanoid robots must autonomously navigate a diverse range of complex terrains with abrupt transitions. While the Vanilla mixture of experts (MoE) framework is theoretically capable of modeling diverse terrain features, in practice, the gating network exhibits nearly uniform expert activations across different terrains, weakening the expert specialization and limiting the model's expressive power. To address this limitation, we introduce CMoE, a novel single-stage reinforcement learning framework that integrates contrastive learning to refine expert activation distributions. By imposing contrastive constraints, CMoE maximizes the consistency of expert activations within the same terrain while minimizing their similarity across different terrains, thereby encouraging experts to specialize in distinct terrain types. We validated our approach on the Unitree G1 humanoid robot through a series of challenging experiments. Results demonstrate that CMoE enables the robot to traverse continuous steps up to 20 cm high and gaps up to 80 cm wide, while achieving robust and natural gait across diverse mixed terrains, surpassing the limits of existing methods. To support further research and foster community development, we release our code publicly.

Training large-scale Mixture-of-Experts (MoE) models is bottlenecked by activation memory and expert-parallel communication, yet FP4 training remains impractical on Hopper-class GPUs without native MXFP4 or NVFP4 support. In this work, we present a training recipe that enables MXFP4 efficiency for MoE models on Hopper architectures without native 4-bit computation support. A central challenge is to integrate FP4 into an existing BF16/FP8 hybrid training pipeline without incurring costly precision round-trips (e.g., FP4 $\leftrightarrow$ BF16 $\leftrightarrow$ FP8). We address this challenge by introducing direct FP8-to-FP4 quantization and de-quantization, together with scaling-aware FP4 row-wise to column-wise conversion, enabling FP4 activations and expert-parallel communication with minimal overhead. Core MoE computations are executed in FP8, while activations and expert-parallel communication are compressed using MXFP4, achieving substantial memory and bandwidth savings without degrading convergence. At the 671B parameter scale, our method achieves end-to-end training performance comparable to strong FP8 baselines, while reducing peak activation memory by 14.8% (11.8 GB) and improving training throughput by 12.5%, from 1157 to 1302 tokens per GPU per second. These results show that FP4 efficiency can be practically realized for large-scale MoE training through careful software-hardware co-design, even without native FP4 Tensor Core support.

Mixture-of-Experts (MoE) architectures have emerged as a powerful paradigm for scaling neural networks while maintaining computational efficiency. However, standard MoE implementations rely on two rigid design assumptions: (1) fixed Top-K routing where exactly K experts are activated per token, and (2) uniform expert allocation across all layers. This paper introduces DynaMoE, a novel MoE framework that relaxes both constraints through dynamic token-level expert activation and layer-wise adaptive capacity allocation. DynaMoE introduces a principled routing mechanism where the number of active experts per token varies based on input complexity. Concurrently, the framework implements six distinct scheduling strategies for distributing expert capacity across network depth, including descending, ascending, pyramid, and wave patterns. We theoretically analyze the expressivity gains of dynamic routing and derive bounds on computational efficiency. Through extensive experiments on MNIST, Fashion-MNIST, CIFAR-10 (image classification), and Recycling-the-Web (language modeling) across multiple model scales, we demonstrate that DynaMoE achieves superior parameter efficiency compared to static baselines. Our key finding is that optimal expert schedules are task- and scale-dependent: descending schedules (concentrating capacity in early layers) outperform uniform baselines on image classification. For language modeling, optimal schedules vary by model size, descending for Tiny, ascending for Small, and uniform for Medium. Furthermore, dynamic routing reduces gradient variance during training, leading to improved convergence stability. DynaMoE establishes a new framework for adaptive computation in neural networks, providing principled guidance for MoE architecture design.

Accurate classification of pediatric central nervous system tumors remains challenging due to histological complexity and limited training data. While pathology foundation models have advanced whole-slide image (WSI) analysis, they often fail to leverage the rich, complementary information found in clinical text and tissue microarchitecture. To this end, we propose PathMoE, an interpretable multimodal framework that integrates H&E slides, pathology reports, and nuclei-level cell graphs via an interaction-aware mixture-of-experts architecture built on state-of-the-art foundation models for each modality. By training specialized experts to capture modality uniqueness, redundancy, and synergy, PathMoE employs an input-dependent gating mechanism that dynamically weights these interactions, providing sample-level interpretability. We evaluate our framework on two dataset-specific classification tasks on an internal pediatric brain tumor dataset (PBT) and external TCGA datasets. PathMoE improves macro-F1 from 0.762 to 0.799 (+0.037) on PBT when integrating WSI, text, and graph modalities; on TCGA, augmenting WSI with graph knowledge improves macro-F1 from 0.668 to 0.709 (+0.041). These results demonstrate significant performance gains over state-of-the-art image-only baselines while revealing the specific modality interactions driving individual predictions. This interpretability is particularly critical for rare tumor subtypes, where transparent model reasoning is essential for clinical trust and diagnostic validation.

Accurate identification of protein active sites at the residue level is crucial for understanding protein function and advancing drug discovery. However, current methods face two critical challenges: vulnerability in single-instance prediction due to sparse training data, and inadequate modality reliability estimation that leads to performance degradation when unreliable modalities dominate fusion processes. To address these challenges, we introduce Multimodal Mixture-of-Experts with Retrieval Augmentation (MERA), the first retrieval-augmented framework for protein active site identification. MERA employs hierarchical multi-expert retrieval that dynamically aggregates contextual information from chain, sequence, and active-site perspectives through residue-level mixture-of-experts gating. To prevent modality degradation, we propose a reliability-aware fusion strategy based on Dempster-Shafer evidence theory that quantifies modality trustworthiness through belief mass functions and learnable discounting coefficients, enabling principled multimodal integration. Extensive experiments on ProTAD-Gen and TS125 datasets demonstrate that MERA achieves state-of-the-art performance, with 90% AUPRC on active site prediction and significant gains on peptide-binding site identification, validating the effectiveness of retrieval-augmented multi-expert modeling and reliability-guided fusion.

Sequential recommendation has rapidly advanced in click-through rate prediction due to its ability to model dynamic user interests. A key challenge, however, lies in modeling long sequences: users often exhibit significant interest shifts, introducing substantial irrelevant or misleading information. Our empirical analysis corroborates this challenge and uncovers a recurring behavioral pattern in long sequences (session hopping): user interests remain stable within short temporal spans (sessions) but shift drastically across sessions and may reappear after multiple sessions. To address this challenge, we propose the Mixture of Sequence (MoS) framework, a model-agnostic MoE approach that achieves accurate predictions by extracting theme-specific and multi-scale subsequences from noisy raw user sequences. First, MoS employs a theme-aware routing mechanism to adaptively learn the latent themes of user sequences and organizes these sequences into multiple coherent subsequences. Each subsequence contains only sessions aligned with a specific theme, thereby effectively filtering out irrelevant or even misleading information introduced by user interest shifts in session hopping. In addition, to alleviate potential information loss, we introduce a multi-scale fusion mechanism, which leverages three types of experts to capture global sequence characteristics, short-term user behaviors, and theme-specific semantic patterns. Together, these two mechanisms endow MoS with the ability to deliver accurate recommendations from multi-faceted and multi-scale perspectives. Experimental results demonstrate that MoS consistently achieves the SOTA performance while introducing fewer FLOPs compared with other MoE counterparts, providing strong evidence of its excellent balance between utility and efficiency. The code is available at https://github.com/xiaolin-cs/MoS.

In recent years, various methods have been proposed for mesh analysis, each offering distinct advantages and often excelling on different object classes. We present a novel Mixture of Experts (MoE) framework designed to harness the complementary strengths of these diverse approaches. We propose a new gate architecture that encourages each expert to specialise in the classes it excels in. Our design is guided by two key ideas: (1) random walks over the mesh surface effectively capture the regions that individual experts attend to, and (2) an attention mechanism that enables the gate to focus on the areas most informative for each expert's decision-making. To further enhance performance, we introduce a dynamic loss balancing scheme that adjusts a trade-off between diversity and similarity losses throughout the training, where diversity prompts expert specialization, and similarity enables knowledge sharing among the experts. Our framework achieves state-of-the-art results in mesh classification, retrieval, and semantic segmentation tasks. Our code is available at: https://github.com/amirbelder/MME-Mixture-of-Mesh-Experts.

Foundation models pre-trained on large-scale datasets demonstrate strong transfer learning capabilities; however, their adaptation to complex multi-label diagnostic tasks-such as comprehensive head CT finding detection-remains understudied. Standard parameter-efficient fine-tuning methods such as LoRA apply uniform adaptations across pathology types, which may limit performance for diverse medical findings. We propose a Mixture of Low-Rank Experts (MoLRE) framework that extends LoRA with multiple specialized low-rank adapters and unsupervised soft routing. This approach enables conditional feature adaptation with less than 0.5% additional parameters and without explicit pathology supervision. We present a comprehensive benchmark of MoLRE across six state-of-the-art medical imaging foundation models spanning 2D and 3D architectures, general-domain, medical-domain, and head CT-specific pretraining, and model sizes ranging from 7M to 431M parameters. Using over 70,000 non-contrast head CT scans with 75 annotated findings-including hemorrhage, infarction, trauma, mass lesions, structural abnormalities, and chronic changes-our experiments demonstrate consistent performance improvements across all models. Gains vary substantially: general-purpose and medical-domain models show the largest improvements (DINOv3-Base: +4.6%; MedGemma: +4.3%), whereas 3D CT-specialized or very large models show more modest gains (+0.2-1.3%). The combination of MoLRE and MedGemma achieves the highest average detection AUC of 0.917. These findings highlight the importance of systematic benchmarking on target clinical tasks, as pretraining domain, architecture, and model scale interact in non-obvious ways.

Large language models (LLMs) achieve remarkable performance on diverse downstream and domain-specific tasks via parameter-efficient fine-tuning (PEFT). However, existing PEFT methods, particularly MoE-LoRA architectures, suffer from limited parameter efficiency and coarse-grained adaptation due to the proliferation of LoRA experts and instance-level routing. To address these issues, we propose Core Space Mixture of LoRA (CoMoL), a novel MoE-LoRA framework that incorporates expert diversity, parameter efficiency, and fine-grained adaptation. Specifically, CoMoL introduces two key components: core space experts and core space routing. Core space experts store each expert in a compact core matrix, preserving diversity while controlling parameter growth. Core space routing dynamically selects and activates the appropriate core experts for each token, enabling fine-grained, input-adaptive routing. Activated core experts are then merged via a soft-merging strategy into a single core expert, which is combined with a shared LoRA to form a specialized LoRA module. Besides, the routing network is projected into the same low-rank space as the LoRA matrices, further reducing parameter overhead without compromising expressiveness. Extensive experiments demonstrate that CoMoL retains the adaptability of MoE-LoRA architectures while achieving parameter efficiency comparable to standard LoRA, consistently outperforming existing methods across multiple tasks.

Mobile Agents can autonomously execute user instructions, which requires hybrid-capabilities reasoning, including screen summary, subtask planning, action decision and action function. However, existing agents struggle to achieve both decoupled enhancement and balanced integration of these capabilities. To address these challenges, we propose Channel-of-Mobile-Experts (CoME), a novel agent architecture consisting of four distinct experts, each aligned with a specific reasoning stage, CoME activates the corresponding expert to generate output tokens in each reasoning stage via output-oriented activation. To empower CoME with hybrid-capabilities reasoning, we introduce a progressive training strategy: Expert-FT enables decoupling and enhancement of different experts' capability; Router-FT aligns expert activation with the different reasoning stage; CoT-FT facilitates seamless collaboration and balanced optimization across multiple capabilities. To mitigate error propagation in hybrid-capabilities reasoning, we propose InfoGain-Driven DPO (Info-DPO), which uses information gain to evaluate the contribution of each intermediate step, thereby guiding CoME toward more informative reasoning. Comprehensive experiments show that CoME outperforms dense mobile agents and MoE methods on both AITZ and AMEX datasets.

Emotion Recognition in Conversations (ERC) presents unique challenges, requiring models to capture the temporal flow of multi-turn dialogues and to effectively integrate cues from multiple modalities. We propose Mixture of Speech-Text Experts for Recognition of Emotions (MiSTER-E), a modular Mixture-of-Experts (MoE) framework designed to decouple two core challenges in ERC: modality-specific context modeling and multimodal information fusion. MiSTER-E leverages large language models (LLMs) fine-tuned for both speech and text to provide rich utterance-level embeddings, which are then enhanced through a convolutional-recurrent context modeling layer. The system integrates predictions from three experts-speech-only, text-only, and cross-modal-using a learned gating mechanism that dynamically weighs their outputs. To further encourage consistency and alignment across modalities, we introduce a supervised contrastive loss between paired speech-text representations and a KL-divergence-based regulariza-tion across expert predictions. Importantly, MiSTER-E does not rely on speaker identity at any stage. Experiments on three benchmark datasets-IEMOCAP, MELD, and MOSI-show that our proposal achieves 70.9%, 69.5%, and 87.9% weighted F1-scores respectively, outperforming several baseline speech-text ERC systems. We also provide various ablations to highlight the contributions made in the proposed approach.

Parameter-efficient fine-tuning has demonstrated promising results across various visual adaptation tasks, such as classification and segmentation. Typically, prompt tuning techniques have harnessed knowledge from a single pre-trained model, whether from a general or a specialized medical domain. However, this approach typically overlooks the potential synergies that could arise from integrating diverse domain knowledge within the same tuning process. In this work, we propose a novel Mixture-of-Experts prompt tuning method called pMoE, which leverages the strengths of multiple expert domains through expert-specialized prompt tokens and the learnable dispatcher, effectively combining their expertise in a unified model framework. Our pMoE introduces expert-specific prompt tokens and utilizes a dynamic token dispatching mechanism at various prompt layers to optimize the contribution of each domain expert during the adaptation phase. By incorporating both domain knowledge from diverse experts, the proposed pMoE significantly enhances the model's versatility and applicability to a broad spectrum of tasks. We conduct extensive experiments across 47 adaptation tasks, including both classification and segmentation in general and medical domains. The results demonstrate that our pMoE not only achieves superior performance with a large margin of improvements but also offers an optimal trade-off between computational efficiency and adaptation effectiveness compared to existing methods.

Neural operators have emerged as an efficient paradigm for solving PDEs, overcoming the limitations of traditional numerical methods and significantly improving computational efficiency. However, due to the diversity and complexity of PDE systems, existing neural operators typically rely on a single network architecture, which limits their capacity to fully capture heterogeneous features and complex system dependencies. This constraint poses a bottleneck for large-scale PDE pre-training based on neural operators. To address these challenges, we propose a large-scale PDE pre-trained neural operator based on a nested Mixture-of-Experts (MoE) framework. In particular, the image-level MoE is designed to capture global dependencies, while the token-level Sub-MoE focuses on local dependencies. Our model can selectively activate the most suitable expert networks for a given input, thereby enhancing generalization and transferability. We conduct large-scale pre-training on twelve PDE datasets from diverse sources and successfully transfer the model to downstream tasks. Extensive experiments demonstrate the effectiveness of our approach.

We propose Excitation, a novel optimization framework designed to accelerate learning in sparse architectures such as Mixture-of-Experts (MoEs). Unlike traditional optimizers that treat all parameters uniformly, Excitation dynamically modulates updates using batch-level expert utilization. It introduces a competitive update dynamic that amplifies updates to highly-utilized experts and can selectively suppress low-utilization ones, effectively sharpening routing specialization. Notably, we identify a phenomenon of "structural confusion" in deep MoEs, where standard optimizers fail to establish functional signal paths; Excitation acts as a specialization catalyst, "rescuing" these models and enabling stable training where baselines remain trapped. Excitation is optimizer-, domain-, and model-agnostic, requires minimal integration effort, and introduces neither additional per-parameter optimizer state nor learnable parameters, making it highly viable for memory-constrained settings. Across language and vision tasks, Excitation consistently improves convergence speed and final performance in MoE models, indicating that active update modulation is a key mechanism for effective conditional computation.

Multimodal recommenders combine collaborative behavior with visual and textual item evidence, whose usefulness varies across user-item interactions. Independently trained source-specific diagnostic probes partition held-out interactions into behavior-, appearance-, semantics-, and mixed-evidence regimes across five benchmarks, within which capacity-matched fixed fusion rules exhibit systematic regime-dependent performance crossovers. This diagnostic observation motivates MAGNET, a multimodal graph recommender with two core mechanisms. First, a calibrated expert bank organizes trainable experts by anchor source (behavior, appearance, or semantics) and fusion family (dominant, balanced, or complementary), while an interaction-conditioned router selects a sparse composition using all three evidence sources. Second, an entropy-triggered, coverage-aware progressive schedule decouples population-level routing-mass coverage from per-instance decisiveness, transitioning from broad routing exploration to confident specialization once sufficient coverage is sustained, while preserving coverage thereafter. MAGNET separately encodes the observed interaction graph and a filtered content-induced structural view, applying cross-view alignment after independent propagation. We evaluate MAGNET on four Amazon domains and the non-Amazon MicroLens-100K benchmark, including tail-item and low-history warm-start user evaluation alongside matched expert-design controls. Under the shared-feature, fixed-split protocol, MAGNET-DV exceeds the strongest protocol-compatible non-MAGNET baseline in every reported main-table cell, supported by seed-wise difference tests. Both the fixed and KL-anchored structured variants achieve higher five-seed mean NDCG@20 than matched homogeneous and free-mixture alternatives across all five datasets. MAGNET supports route-level diagnostics through explicit expert semantics.

Traditional Mixture-of-Experts (MoE) training typically proceeds without any structural priors, effectively requiring the model to simultaneously train expert weights while searching for an optimal routing policy within a vast combinatorial space. This entanglement often leads to sluggish convergence and training instabilities. This paper introduces Grouter, a preemptive routing method that by distilling high-quality structures from fully-trained MoE models and serving as a fixed router for target models. By decoupling structural optimization from weight updates, Grouter significantly accelerates both the speed and quality of model convergence. To ensure the framework's versatility, we also introduce expert folding to adapt Grouter across varying model configurations and expert tuning to rebalance workloads across different data distributions. Furthermore, by leveraging the structural priors provided by preemptive routing, we can implement targeted optimizations to further enhance training throughput. Experiments demonstrate that Grouter achieves superior performance and efficiency which boosts pre-training data utilization by 4.28x and achieves up to 33.5% throughput acceleration, establishing preemptive routing as a fundamental paradigm for scalable MoE training. We publicly release our code and pretrained Grouter checkpoints at https://github.com/JimmyAwoe/Grouter.

Mixture-of-Experts (MoE) offers flexible graph reasoning by combining multiple views of a graph through a learned router. We investigate routing-aware explanations for MoE graph models in malware detection using control flow graphs (CFGs). Our architecture builds diversity at two levels. At the node level, each layer computes multiple neighborhood statistics and fuses them with an MLP, guided by a degree reweighting factor rho and a pooling choice lambda in {mean, std, max}, producing distinct node representations that capture complementary structural cues in CFGs. At the readout level, six experts, each tied to a specific (rho, lambda) view, output graph-level logits that the router weights into a final prediction. Post-hoc explanations are generated with edge-level attributions per expert and aggregated using the router gates so the rationale reflects both what each expert highlights and how strongly it is selected. Evaluated against single-expert GNN baselines such as GCN, GIN, and GAT on the same CFG dataset, the proposed MoE achieves strong detection accuracy while yielding stable, faithful attributions under sparsity-based perturbations. The results indicate that making the router explicit and combining multi-statistic node encoding with expert-level diversity can improve the transparency of MoE decisions for malware analysis.

Reinforcement learning (RL) has equipped LLM agents with a strong ability to solve complex tasks. However, existing RL methods normally use a single policy network, causing simplicity bias where simple tasks occupy most parameters and dominate gradient updates, leaving insufficient capacity for complex tasks. A plausible remedy could be employing the Mixture-of-Experts (MoE) architecture in the policy network, as MoE allows different parameters (experts) to specialize in different tasks, preventing simple tasks from dominating all parameters. However, a key limitation of traditional MoE is its token-level routing, where the router assigns each token to specialized experts, which fragments phase-consistent patterns into scattered expert assignments and thus undermines expert specialization. In this paper, we propose Phase-Aware Mixture of Experts (PA-MoE). It first features a lightweight phase router that learns latent phase boundaries directly from the RL objective without pre-defining phase categories. Then, the phase router allocates temporally consistent assignments to the same expert, allowing experts to preserve phase-specific expertise. Experimental results demonstrate the effectiveness of our proposed PA-MoE.

Mixture-of-Experts (MoE) architectures decompose prediction tasks into specialized expert sub-networks selected by a gating mechanism. This letter adopts a communication-theoretic view of MoE gating, modeling the gate as a stochastic channel operating under a finite information rate. Within an information-theoretic learning framework, {we specialize a mutual-information generalization bound and develop a rate-distortion characterization $D(R_g)$ of finite-rate gating, where $R_g:=I(X; T)$, yielding (under a standard empirical rate-distortion optimality condition) $\mathbb{E}[R(W)] \le D(R_g)+δ_m+\sqrt{(2/m)\, I(S; W)}$. }The analysis yields capacity-aware limits for communication-constrained MoE systems, and numerical simulations on synthetic multi-expert models empirically confirm the predicted trade-offs between gating rate, expressivity, and generalization.

Vision-Language-Action (VLA) models for autonomous driving increasingly adopt generative planners trained with imitation learning followed by reinforcement learning. Diffusion-based planners suffer from modality alignment difficulties, low training efficiency, and limited generalization. Token-based planners are plagued by cumulative causal errors and irreversible decoding. In summary, the two dominant paradigms exhibit complementary strengths and weaknesses. In this paper, we propose DriveFine, a masked diffusion VLA model that combines flexible decoding with self-correction capabilities. In particular, we design a novel plug-and-play block-MoE, which seamlessly injects a refinement expert on top of the generation expert. By enabling explicit expert selection during inference and gradient blocking during training, the two experts are fully decoupled, preserving the foundational capabilities and generic patterns of the pretrained weights, which highlights the flexibility and extensibility of the block-MoE design. Furthermore, we design a hybrid reinforcement learning strategy that encourages effective exploration of refinement expert while maintaining training stability. Extensive experiments on NAVSIM v1, v2, and Navhard benchmarks demonstrate that DriveFine exhibits strong efficacy and robustness. The code will be released at https://github.com/MSunDYY/DriveFine.

Mixture-of-Experts (MoE) architectures are evolving towards finer granularity to improve parameter efficiency. However, existing MoE designs face an inherent trade-off between the granularity of expert specialization and hardware execution efficiency. We propose OmniMoE, a system-algorithm co-designed framework that pushes expert granularity to its logical extreme. OmniMoE introduces vector-level Atomic Experts, enabling scalable routing and execution within a single MoE layer, while retaining a shared dense MLP branch for general-purpose processing. Although this atomic design maximizes capacity, it poses severe challenges for routing complexity and memory access. To address these, OmniMoE adopts a system-algorithm co-design: (i) a Cartesian Product Router that decomposes the massive index space to reduce routing complexity from O(N) to O(sqrt(N)); and (ii) Expert-Centric Scheduling that inverts the execution order to turn scattered, memory-bound lookups into efficient dense matrix operations. Validated on seven benchmarks, OmniMoE (with 1.7B active parameters) achieves 50.9% zero-shot accuracy across seven benchmarks, outperforming coarse-grained (e.g., DeepSeekMoE) and fine-grained (e.g., PEER) baselines. Crucially, OmniMoE reduces inference latency from 73ms to 6.7ms (a 10.9-fold speedup) compared to PEER, demonstrating that massive-scale fine-grained MoE can be fast and accurate. Our code is open-sourced at https://github.com/flash-algo/omni-moe.

Real-world time series exhibit complex and evolving dynamics, making accurate forecasting extremely challenging. Recent multi-modal forecasting methods leverage textual information such as news reports to improve prediction, but most rely on token-level fusion that mixes temporal patches with language tokens in a shared embedding space. However, such fusion can be ill-suited when high-quality time-text pairs are scarce and when time series exhibit substantial variation in characteristics, thus complicating cross-modal alignment. In parallel, mixture-of-experts (MoE) architectures have proven effective for both time series modeling and multi-modal learning, yet many existing MoE-based modality integration methods still depend on token-level fusion. To address this, we propose Expert Modulation, a new mechanism for multi-modal time series prediction that conditions both routing and expert computation on textual signals, enabling direct and efficient cross-modal control over expert behavior. Through theoretical analysis and experiments, our proposed method demonstrates strong improvements in multi-modal time series prediction. The current code implementation is available at https://github.com/BruceZhangReve/MoME

Mixture-of-Experts (MoE) models scale neural networks by conditionally activating a small subset of experts, where the router plays a central role in determining expert specialization and overall model performance. However, many modern MoE systems still adopt linear routers in raw high-dimensional representation spaces, where representation mismatch, angular concentration, and scale-sensitive scoring can jointly undermine routing discriminability and stable expert specialization. In this work, we propose Low-rank & Lipschitz-controlled Routing (L2R), a unified routing framework that reshapes both the routing space and scoring geometry. L2R performs expert assignment in a shared low-rank latent routing space and introduces Saturated Inner-Product Scoring (SIPS) to explicitly control the Lipschitz behavior of routing functions, yielding smoother and more stable routing geometry. In addition, L2R incorporates a parameter-efficient multi-anchor routing mechanism to enhance expert expressiveness. Experiments on an OLMoE-based language MoE model and a ViT-based ImageNet setting show improved overall performance in both domains; OLMoE diagnostics further show improved routing geometry and expert discrimination.