Copertina

Why Smaller Is Slower? Dimensional Misalignment in Compressed LLMs

Fuente: arXiv

Salvato in:

Dettagli Bibliografici
Autori principali:	Xin, Jihao, Lyu, Tian, Pan, Qilong, Wang, Kesen, Canini, Marco
Natura:	Preprint
Pubblicazione:	2026
Soggetti:	Distributed, Parallel, and Cluster Computing Artificial Intelligence
Accesso online:	Acceder al recurso
Tags:	Aggiungi Tag Nessun Tag, puoi essere il primo ad aggiungerne!!

Documenti analoghi

Quantize Once, Train Fast: Allreduce-Compatible Compression with Provable Guarantees
di: Xin, Jihao, et al.
Pubblicazione: (2023)

TACO: Efficient Communication Compression of Intermediate Tensors for Scalable Tensor-Parallel LLM Training
di: Liu, Man, et al.
Pubblicazione: (2026)

FilFL: Client Filtering for Optimized Client Participation in Federated Learning
di: Fourati, Fares, et al.
Pubblicazione: (2023)

Why Does the LLM Stop Computing: An Empirical Study of User-Reported Failures in Open-Source LLMs
di: Yu, Guangba, et al.
Pubblicazione: (2026)

NeurLZ: An Online Neural Learning-Based Method to Enhance Scientific Lossy Compression
di: Jia, Wenqi, et al.
Pubblicazione: (2024)

ModTrans: Translating Real-world Models for Distributed Training Simulator
di: Lyu, Yi
Pubblicazione: (2026)

Why Should the Server Do It All?: A Scalable, Versatile, and Model-Agnostic Framework for Server-Light DNN Inference over Massively Distributed Clients via Training-Free Intermediate Feature Compression
di: Sung, Mingyu, et al.
Pubblicazione: (2025)

IoT-MCP: Bridging LLMs and IoT Systems Through Model Context Protocol
di: Yang, Ningyuan, et al.
Pubblicazione: (2025)

UCCL-Zip: Lossless Compression Supercharged GPU Communication
di: Ma, Shuang, et al.
Pubblicazione: (2026)

Failure-Resilient Distributed Inference with Model Compression over Heterogeneous Edge Devices
di: Wang, Li, et al.
Pubblicazione: (2024)

GWLZ: A Group-wise Learning-based Lossy Compression Framework for Scientific Data
di: Jia, Wenqi, et al.
Pubblicazione: (2024)

Training LLMs with Fault Tolerant HSDP on 100,000 GPUs
di: Salpekar, Omkar, et al.
Pubblicazione: (2026)

ELANA: A Simple Energy and Latency Analyzer for LLMs
di: Chiang, Hung-Yueh, et al.
Pubblicazione: (2025)

Why Do AI Agents Systematically Fail at Cloud Root Cause Analysis?
di: Kim, Taeyoon, et al.
Pubblicazione: (2026)

PacTrain: Pruning and Adaptive Sparse Gradient Compression for Efficient Collective Communication in Distributed Deep Learning
di: Wang, Yisu, et al.
Pubblicazione: (2025)

SMART: When is it Actually Worth Expanding a Speculative Tree?
di: Wang, Lifu, et al.
Pubblicazione: (2026)

Zipage: Maintain High Request Concurrency for LLM Reasoning through Compressed PagedAttention
di: Liao, Mengqi, et al.
Pubblicazione: (2026)

Accelerating Large Language Model Training with Hybrid GPU-based Compression
di: Xu, Lang, et al.
Pubblicazione: (2024)

Byzantine-Robust and Communication-Efficient Distributed Training: Compressive and Cyclic Gradient Coding
di: Li, Chengxi, et al.
Pubblicazione: (2026)

KVComp: A High-Performance, LLM-Aware, Lossy Compression Framework for KV Cache
di: Jiang, Bo, et al.
Pubblicazione: (2025)

A Model Aware AIGC Task Offloading Algorithm in IIoT Edge Computing
di: Wang, Xin, et al.
Pubblicazione: (2025)

Smaller, Smarter, Closer: The Edge of Collaborative Generative AI
di: Morabito, Roberto, et al.
Pubblicazione: (2025)

PackKV: Reducing KV Cache Memory Footprint through LLM-Aware Lossy Compression
di: Jiang, Bo, et al.
Pubblicazione: (2025)

FairBatching: Fairness-Aware Batch Formation for LLM Inference
di: Lyu, Hongtao, et al.
Pubblicazione: (2025)

An Evaluation of LLMs Inference on Popular Single-board Computers
di: Tung, et al.
Pubblicazione: (2025)

Balanced and Elastic End-to-end Training of Dynamic LLMs
di: Wahib, Mohamed, et al.
Pubblicazione: (2025)

FlashRecovery: Fast and Low-Cost Recovery from Failures for Large-Scale Training of LLMs
di: Zhang, Haijun, et al.
Pubblicazione: (2025)

SwizzlePerf: Hardware-Aware LLMs for GPU Kernel Performance Optimization
di: Tschand, Arya, et al.
Pubblicazione: (2025)

High-Dimensional Data Processing: Benchmarking Machine Learning and Deep Learning Architectures in Local and Distributed Environments
di: Rodriguez, Julian, et al.
Pubblicazione: (2025)

xDiT: an Inference Engine for Diffusion Transformers (DiTs) with Massive Parallelism
di: Fang, Jiarui, et al.
Pubblicazione: (2024)

Efficient MoE Inference with Fine-Grained Scheduling of Disaggregated Expert Parallelism
di: Pan, Xinglin, et al.
Pubblicazione: (2025)

Design a Win-Win Strategy That Is Fair to Both Service Providers and Tasks When Rejection Is Not an Option
di: Trabelsi, Yohai, et al.
Pubblicazione: (2024)

Research on Model Parallelism and Data Parallelism Optimization Methods in Large Language Model-Based Recommendation Systems
di: Yang, Haowei, et al.
Pubblicazione: (2025)

FusionLLM: A Decentralized LLM Training System on Geo-distributed GPUs with Adaptive Compression
di: Tang, Zhenheng, et al.
Pubblicazione: (2024)

ScaleSim: Serving Large-Scale Multi-Agent Simulation with Invocation Distance-Based Memory Management
di: Pan, Zaifeng, et al.
Pubblicazione: (2026)

Marconi: Prefix Caching for the Era of Hybrid LLMs
di: Pan, Rui, et al.
Pubblicazione: (2024)

Flashback: Understanding and Mitigating Forgetting in Federated Learning
di: Aljahdali, Mohammed, et al.
Pubblicazione: (2024)

FreeRide: Harvesting Bubbles in Pipeline Parallelism
di: Zhang, Jiashu, et al.
Pubblicazione: (2024)

Domain-Adaptive Model Merging Across Disconnected Modes
di: Liu, Junming, et al.
Pubblicazione: (2026)

ClusterRCA: An End-to-End Approach for Network Fault Localization and Classification for HPC System
di: Sun, Yongqian, et al.
Pubblicazione: (2025)