Sparsity-Controllable Dynamic Top-p MoE for Large Foundation Model Pre-training

• URL: http://arxiv.org/abs/2512.13996v1
• Date: Tue, 16 Dec 2025 01:28:57 GMT
• Title: Sparsity-Controllable Dynamic Top-p MoE for Large Foundation Model Pre-training
• Authors: Can Jin, Hongwu Peng, Mingcan Xiang, Qixin Zhang, Xiangchi Yuan, Amit Hasan, Ohiremen Dibua, Yifan Gong, Yan Kang, Dimitris N. Metaxas,
• Abstract summary: DTop-p is a sparsity-controllable dynamic Top-p routing mechanism.<n>We show that DTop-p consistently outperforms both Top-k and fixed-threshold Top-p baselines.<n>DTop-p exhibits strong scaling properties with respect to expert granularity, expert capacity, model size, and dataset size.
• Score: 30.589225478300023
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Sparse Mixture-of-Experts (MoE) architectures effectively scale model capacity by activating only a subset of experts for each input token. However, the standard Top-k routing strategy imposes a uniform sparsity pattern that ignores the varying difficulty of tokens. While Top-p routing offers a flexible alternative, existing implementations typically rely on a fixed global probability threshold, which results in uncontrolled computational costs and sensitivity to hyperparameter selection. In this paper, we propose DTop-p MoE, a sparsity-controllable dynamic Top-p routing mechanism. To resolve the challenge of optimizing a non-differentiable threshold, we utilize a Proportional-Integral (PI) Controller that dynamically adjusts the probability threshold to align the running activated-expert sparsity with a specified target. Furthermore, we introduce a dynamic routing normalization mechanism that adapts layer-wise routing logits, allowing different layers to learn distinct expert-selection patterns while utilizing a global probability threshold. Extensive experiments on Large Language Models and Diffusion Transformers demonstrate that DTop-p consistently outperforms both Top-k and fixed-threshold Top-p baselines. Our analysis confirms that DTop-p maintains precise control over the number of activated experts while adaptively allocating resources across different tokens and layers. Furthermore, DTop-p exhibits strong scaling properties with respect to expert granularity, expert capacity, model size, and dataset size, offering a robust framework for large-scale MoE pre-training.

Related papers

• DynaMoE: Dynamic Token-Level Expert Activation with Layer-Wise Adaptive Capacity for Mixture-of-Experts Neural Networks [0.0]
  Mixture-of-Experts (MoE) architectures have emerged as a powerful paradigm for scaling neural networks while maintaining computational efficiency.<n>This paper introduces DynaMoE, a novel MoE framework that relaxes both constraints through dynamic token-level expert activation and layer-wise adaptive capacity allocation.
  arXiv  Detail & Related papers  (2026-03-02T10:25:56Z)
• AR-MOT: Autoregressive Multi-object Tracking [56.09738000988466]
  We propose a novel autoregressive paradigm that formulates MOT as a sequence generation task within a large language model (LLM) framework.<n>This design enables the model to output structured results through flexible sequence construction, without requiring any task-specific heads.<n>To enhance region-level visual perception, we introduce an Object Tokenizer based on a pretrained detector.
  arXiv  Detail & Related papers  (2026-01-05T09:17:28Z)
• Reasoning Palette: Modulating Reasoning via Latent Contextualization for Controllable Exploration for (V)LMs [49.66344956133349]
  Reasoning capacity shapes both inference-time performance and reinforcement learning (RL) training for large (vision-) language models.<n>This paper proposes Reasoning Palette, a novel latent-modulation framework that endows the model with a latent variable for strategic contextualization.
  arXiv  Detail & Related papers  (2025-12-19T03:32:53Z)
• An Integrated Fusion Framework for Ensemble Learning Leveraging Gradient Boosting and Fuzzy Rule-Based Models [59.13182819190547]
  Fuzzy rule-based models excel in interpretability and have seen widespread application across diverse fields.<n>They face challenges such as complex design specifications and scalability issues with large datasets.<n>This paper proposes an Integrated Fusion Framework that merges the strengths of both paradigms to enhance model performance and interpretability.
  arXiv  Detail & Related papers  (2025-11-11T10:28:23Z)
• TuckA: Hierarchical Compact Tensor Experts for Efficient Fine-Tuning [83.93651411533533]
  We introduce Tucker Adaptation (TuckA), a method with four key properties.<n>We develop an efficient batch-level routing mechanism, which reduces the router's parameter size by a factor of $L$.<n>Experiments on benchmarks in natural language understanding, image classification, and mathematical reasoning speak to the efficacy of TuckA.
  arXiv  Detail & Related papers  (2025-11-10T09:03:16Z)
• Controllable-LPMoE: Adapting to Challenging Object Segmentation via Dynamic Local Priors from Mixture-of-Experts [16.21786310193235]
  We propose a novel dynamic priors-based fine-tuning paradigm with fewer trainable parameters, dubbed Controllable-LPMoE.<n>We construct a lightweight dynamic mixed local priors extractor that captures diverse local priors from input images through heterogeneous convolutions.<n>We also design a bi-directional interaction adapter that employs cosine-aligned deformable attention and channel-oriented adaptive scale enhancement.
  arXiv  Detail & Related papers  (2025-10-24T03:03:59Z)
• LD-MoLE: Learnable Dynamic Routing for Mixture of LoRA Experts [24.0422448103907]
  We propose LD-MoLE, a Learnable Dynamic routing mechanism for Mixture of LoRA Experts.<n>Our design allows the model to adaptively determine the number of experts to activate for each token at different layers.<n>Our method achieves superior performance, but also demonstrates the ability to learn token-dependent and layer-wise expert allocation.
  arXiv  Detail & Related papers  (2025-09-30T02:38:10Z)
• Expert Race: A Flexible Routing Strategy for Scaling Diffusion Transformer with Mixture of Experts [33.39800923804871]
  We introduce Race-DiT, a novel MoE model for diffusion transformers with a flexible routing strategy, Expert Race.<n>By allowing tokens and experts to compete together and select the top candidates, the model learns to dynamically assign experts to critical tokens.
  arXiv  Detail & Related papers  (2025-03-20T11:45:08Z)
• DiffMoE: Dynamic Token Selection for Scalable Diffusion Transformers [86.5541501589166]
  DiffMoE is a batch-level global token pool that enables experts to access global token distributions during training.<n>It achieves state-of-the-art performance among diffusion models on ImageNet benchmark.<n>The effectiveness of our approach extends beyond class-conditional generation to more challenging tasks such as text-to-image generation.
  arXiv  Detail & Related papers  (2025-03-18T17:57:07Z)
• Unchosen Experts Can Contribute Too: Unleashing MoE Models' Power by Self-Contrast [58.98411447739218]
  Mixture-of-Experts (MoE) has emerged as a prominent architecture for scaling model size while maintaining computational efficiency. We propose Self-Contrast Mixture-of-Experts (SCMoE), a training-free strategy that utilizes unchosen experts in a self-contrast manner during inference. Our method is conceptually simple and computationally lightweight, as it incurs minimal latency compared to greedy decoding.
  arXiv  Detail & Related papers  (2024-05-23T12:45:29Z)
• Trajectory-wise Multiple Choice Learning for Dynamics Generalization in Reinforcement Learning [137.39196753245105]
  We present a new model-based reinforcement learning algorithm that learns a multi-headed dynamics model for dynamics generalization. We incorporate context learning, which encodes dynamics-specific information from past experiences into the context latent vector. Our method exhibits superior zero-shot generalization performance across a variety of control tasks, compared to state-of-the-art RL methods.
  arXiv  Detail & Related papers  (2020-10-26T03:20:42Z)

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.