Spectral-MoE introduces a method for allocating heterogeneous per-layer expert utilization in dense-to-mixture-of-experts (MoE) conversion of large language models. It offers both static and dynamic allocators that learn to distribute a utilization budget across layers, improving performance over uniform allocation in both analytic- and learning-based conversion baselines. Experiments on Llama-2-7B, Llama-3-8B, and Qwen2.5-7B show significant reductions in perplexity and accuracy gains at various FFN utilizations.
arXiv:2606. 01666v1 Announce Type: cross Abstract: The scaling of Large Language Models (LLMs) has driven significant performance gains but created substantial challenges in inference efficiency.
arXiv:2609.36222v1 Announce Type: new Abstract: Large language models are increasingly expensive to serve. In large-scale serving systems, autoregressive decoding is often bottlenecked by transferrin...
MASKerade is a new dense‑to‑MoE training method that learns experts as sparse subnetworks of a frozen pretrained feed‑forward network (FFN) using binary masks. A token‑level router selects which masked FFNs to execute, and both the router and mask scores are jointly optimized while the underlying FFN weights remain unchanged. In experiments on five vision‑language benchmarks with Qwen and Gemma backbones, a configuration of four 2:4 experts with top‑2 routing achieved the best performance among compared baselines, demonstrating that mask learning over frozen weights is a practical alternative for constructing token‑routed MoE experts.
IntBMoE introduces a block‑conditioned mixture‑of‑experts that decouples participation, execution, and materialization by combining dense expert composition with sparse block execution. Each internal layer uses a lightweight hypernetwork to merge all expert bases into a single composed expert, while a router selects only a few blocks per token, keeping compute and memory costs low. Experiments on image classification, language modeling, and sequential recommendation demonstrate consistent performance gains, and the model is deployed in AMap’s generative recommendation system, improving UVCTR by 2.4% in online A/B tests.
arXiv:2607. 00162v1 Announce Type: new Abstract: Parameter-efficient fine-tuning (PEFT) reparameterizes weight updates in a fixed basis: low-rank adapters operate in the spatial domain, while a recent line of spectral methods operates in a fixed Fourier domain.
arXiv:2608. 03457v1 Announce Type: new Abstract: Diffusion language models (dLLMs) offer an alternative to autoregressive (AR) language modeling, yet the scaling behavior of Mixture-of-Experts (MoE) dLLMs remains poorly understood.
The paper introduces Neural Spectral Capacity (NSC), a closed‑form metric derived from the singular‑value spectrum of weight matrices that can be computed solely from a network’s architectural specification. Unlike traditional measures such as #Params and #FLOPs, NSC captures architectural structure (depth, width, head, FFN allocations) and can be evaluated without instantiating the model, data, or gradients. Using a dynamic‑programming solver (NSC‑DP), the authors demonstrate that NSC can efficiently identify architectures that outperform existing training‑free proxies across Transformer and CNN families, and achieve state‑of‑the‑art results in tasks such as WikiText‑103 and commonsense reasoning with LLaMA‑7B. whyItMatters":"NSC provides a fast, architecture‑only proxy that outperforms conventional metrics and training‑free proxies, enabling more effective design and pruning of large models without costly training or data."
ITC-MoE introduces an Importance-guided Token-aware Compression framework for Mixture-of-Experts Diffusion Language Models. It combines Adaptive Tucker Compression, which uses activation and gradient importance to jointly factorize expert weights and allocate ranks, with Token-aware Compensation and Routing that applies low‑rank adjustments to hot tokens and limits expert candidates for cold tokens. The method achieves significant reductions in computation and storage while maintaining generation quality, exemplified by a 30% compression budget that preserves 96.33% accuracy on MultiArith and delivers up to a 7.22× speedup.
arXiv:2609.40127v1 Announce Type: cross Abstract: Modern transformers pair impressive capabilities with substantial memory and compute demands. Low-rank weight factorization reduces both while keepin...
arXiv:2606. 15453v1 Announce Type: cross Abstract: Mixture-of-Experts (MoE) based large language models (LLMs), such as Qwen and DeepSeek, have recently emerged as an effective approach to improving model capacity without proportionally increasing computational cost.
arXiv:2606. 04980v1 Announce Type: new Abstract: Mixture-of-Experts (MoE) architectures scale model capacity through sparse expert activation, but their deployment remains memory-bound because all expert weights must reside in memory.
MetaNet is a support‑set controller that predicts, for each layer of a Mixture‑of‑Experts model, an expert‑retention threshold and a bounded routing bias while keeping the backbone, experts, and router frozen. On DeepSeek‑MoE‑16B‑Chat, MetaNet offers a tunable trade‑off between accuracy and expert activation: a conservative setting activates 3.61 experts on average (40% fewer than a fixed k=6) with comparable MMLU accuracy, whereas an aggressive setting activates only 2.28 experts (62% fewer) with a modest accuracy drop. The MMLU‑trained controller also transfers to C‑Eval, activating 2.90 experts on average (52% fewer than fixed k=6) at 0.386 accuracy.