This paper proposes a four-layer technical architecture for large model inference optimization, including Multi-model Fusion, Model Optimization, Compute-Model Fusion, and Compute-Network-Model Fusion…

This paper introduces a distribution-aware framework for modeling and benchmarking Mixture-of-Experts (MoE) inference, showing that the best fused-MoE kernel changes with routing skew and token count,…

The paper introduces TrojanMerge, a framework demonstrating that model merging can be exploited to systematically compromise the safety alignment of multiple individually safe LLMs.

Taurus is a single-machine system for efficient Graph Neural Network (GNN) inference on large-scale graphs that do not fit in RAM, using source-centric broadcasts and a pipelined GPU-CPU-SSD hierarchy…

The paper proposes MITL, an MsFEM-inspired transfer learning strategy for CNN-based reduced-order models, enabling efficient and adaptable approximation of multiscale systems with minimal retraining.

This paper proposes a new router redesign for Mixture-of-Experts models using Manifold Power Iteration to align router rows with the principal singular directions of associated experts.

The paper proposes a fine-grained approach to overlap expert compute with the second all-to-all communication in Mixture-of-Experts (MoE) models, improving distributed MoE execution efficiency on mult…

The paper proposes Task-Aware Coactivation Grouping (TACG) to significantly reduce communication costs in multi-task MoE inference by grouping experts based on task-specific co-activation patterns, ou…

The paper introduces TN-SHAP-G, a novel framework that uses graph-structured tensor networks to efficiently approximate and compute Shapley values and interaction indices for black-box models, overcom…

This paper presents a memory-efficient training stack for Mixture-of-Experts (MoE) models, combining and specializing parallelism techniques for maximal efficiency.

The paper proposes DAG-MoE, a novel sparse Mixture-of-Experts framework that replaces standard weighted-sum aggregation with structural aggregation to enhance model performance and enable multi-step r…

ResMerge proposes a residual-based spectral merging framework that improves the combination of multiple reinforcement learning (RL) expert models by stabilizing the aggregation process using a residua…

The paper systematically analyzes the benefits and limits of Attention-FFN Disaggregation (AFD) for Mixture-of-Experts (MoE) LLM serving, demonstrating that AFD is crucial for achieving high throughpu…

Hawkeye is a system that allows perfect, precision-preserving reproduction of GPU-level matrix multiplication operations on a CPU, enabling efficient and trustworthy third-party auditing of machine le…

This paper presents Roomie, an architecture for orchestrating model serving in multi-model scenarios on GPUs, reducing SLO violations by predicting and avoiding kernel-level interference.

This paper introduces a mechanistic neuronal network model for multilayer learning, offering biological insights and an alternative to backpropagation.

The elasticAI.explorer is an extensible, unified Python framework that simplifies hardware-aware Neural Architecture Search (NAS) by decoupling search space definition from model implementation and de…

The paper presents a framework for model selection and parameter estimation using large language models and neural simulation-based inference.

The paper presents a field study on deploying large inference workloads on a non-GPU AI accelerator and identifies eight categories of limitations.