RAZOR, a training-free pruning method based on consensus residuals, the deviations of expert outputs from their original weighted mixture, gives the exact output change for a single deletion under survivor renormalization and router refill.
Abstract
Mixture-of-experts (MoE) models activate only a few experts per token but store the entire expert pool. Pruning this pool requires identifying experts whose removal preserves model behavior. Routing frequency and output magnitude do not fully describe deletion damage, which also depends on how the surviving and replacement experts compensate for the removed output. We introduce RAZOR, a training-free pruning method based on consensus residuals, the deviations of expert outputs from their original weighted mixture. At a fixed layer input, these residuals give the exact output change for a single deletion under survivor renormalization and router refill. RAZOR aggregates this damage by conditional root mean square and selects experts under a layerwise budget using forward computation alone, without gradients, subset search, or recovery training. Against frequency, activation-norm, and REAP baselines on GLM-4.7-Flash and Qwen3.6-35B-A3B at 25% and 50% expert removal, it attains the highest macro average over nine reasoning-intensive tasks in all four model-budget settings, gaining 2.12-5.59 points over REAP and lowering reverse KL in all four. On DeepSeek-V4-Flash-0731 and Hy3, it also achieves the highest macro average among the three residual criteria. Local exactness does not guarantee better joint pruning. Generation analyses show changes in diversity, formatting, and termination despite higher task scores.
OptEP is developed, a scalable continuous relaxation that directly optimizes differentiable expert-selection variables, followed by a lightweight reconstruction-based update of the retained expert weights, demonstrating that reconstruction-driven expert selection and weight adaptation can preserve substantially more of...
Jelena Markovic-Voronov, P. Prastakos, Kayhan Behdin et al.· 0 citations
Mixture-of-Experts (MoE) models enable parameter scaling with limited per-token computation by activating only a small subset of experts for each token, but deploying them still requires loading the complete expert pool into memory. Structured expert pruning can effectively reduce the memory usage by removing experts....
Yu-Shuai Sun, Zi-Kun Zhou, Lin Gao et al.· 0 citations
Distributionally Robust MoE Training (DRMoET), a drop-in objective that treats layer-wise experts as endogenous robustness groups and optimizes high-loss routing outcomes rather than merely equalizing traffic, is proposed.
Mixture-of-Experts (MoE) language models suffer from large parameter counts, which create a significant memory bottleneck. Expert pruning is the most direct approach for reducing this parameter count, yet existing methods make pruning decisions for each expert independently, and assume experts'contributions are purely...
Alex M. Tseng, Prannay Kaul, L. Zancato et al.· arXiv.org· 2 citations
SlimWise is presented, a serving framework that tailors the expert pool to each inference phase that improves decode throughput by up to 1.81x at 50% expert pruning with minimal accuracy loss and shows that benchmark accuracy can conceal substantial pruning-induced changes in generation length.
Gunho Park, Kyoungho Jeun, Juntaek Oh et al.· arXiv.org· 1 citation
The ablations analyse why Foil works and turn the findings into design guidance for looped MoE: the returns of looping and of widening the expert layers amplify each other, routing confidence tracks healthy expert use better than load balance, and a sparse looped MoE should therefore use more experts per layer and more...
By studying the computational foundations of language learning, Associate Professor Jacob Andreas builds intelligent systems that can communicate effectively with humans.
MIT News · Artificial Intelligence· news.mit.eduAug 27, 2026
A new machine-learning framework aims to improve the success rate of computational protein design while moving away from results that reproduce sequences found in nature.