This paper proposes Hierarchical Block-Local Learning (HBLL), a framework for training deep neural networks without full end-to-end backpropagation, achieving $\mathcal{O}(\log N)$ parallel time compl…

This paper compares and optimizes CUDA strategies for a shallow neural network, achieving a 1.41x speedup on a large dataset.

This paper proposes DPIFrame, a dual parallelizable framework for accelerating Click-through rate (CTR) model inference on GPU, achieving state-of-the-art inference performance with significant speedu…

LALE introduces a novel lightweight architecture that efficiently combines local convolutional features and global transformer context for land-cover segmentation, achieving superior efficiency and pe…

This paper proposes a new reinforcement learning approach for learning-to-rank that focuses on variance reduction and GPU computation, enabling seamless integration with auto-differentiation and signi…

肖代替了视觉令牌的永久删除，通过可恢复的路由来改进视觉语言模型的性能

Physical AI inference (batch-1 decode) is primarily memory-bandwidth-bound, but the observed latency gap between fast and slow GPUs is not solely due to memory bandwidth, as launch-side overheads beco…

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

ANNLib is a library for Approximate Nearest Neighbor Search (ANNS) providing high performance and flexible functionality using graph-based algorithms and data structures.

This paper develops specialized, I/O-aware GPU kernels for common GNN layer types, achieving significant speedups and memory reductions compared to existing frameworks.

CART introduces a parameter-efficient recurrent transformer architecture that reuses a core block multiple times, but its performance does not surpass a dense baseline, suggesting that weight sharing…

The paper introduces NaRA, a noise-aware LoRA technique that dynamically adapts fine-tuning parameters based on the noise level during diffusion, significantly improving the performance of Diffusion L…

The paper proposes two novel multi-column RBFN architectures, MC-PSO and MC-APSO, that combine parallel RBFN structures with swarm optimization to significantly outperform existing methods in accuracy…

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…

BudgetDraft introduces an acceptance-aware multi-view training method that trains a sparse-KV speculative decoder to maintain high acceptance rates across varying context lengths and sparsity levels,…