No GPU left behind: Unlocking Efficiency with Co-located vLLM in TRL

An Efficient Heterogeneous Co-Design for Fine-Tuning on a Single GPU

arXiv:2603. 16428v2 Announce Type: replace-cross Abstract: Fine-tuning Large Language Models (LLMs) has become essential for domain adaptation, but its memory-intensive property exceeds the capabilities of most GPUs.

From Zero to GPU: A Guide to Building and Scaling Production-Ready CUDA Kernels

CTA-Pipelining: A Latency-Oriented Spatial Scaling Method for Multi-GPU Systems

arXiv:2607. 07862v1 Announce Type: cross Abstract: The evolution of compute infrastructure has transformed multi-GPU systems into tightly integrated shared-memory structures.

DynaResize: Runtime GPU Reallocation for Disaggregated LLM Post-Training

arXiv:2607. 22614v1 Announce Type: new Abstract: RL-based LLM post-training increasingly disaggregates Rollout and Training across separate GPU resources, but static GPU partitioning suffers from severe pipeline bubbles under long-tail rollout latency.

CUCo: An Agentic Framework for Compute and Communication Co-design

arXiv:2603. 02376v2 Announce Type: replace-cross Abstract: Computation and communication in distributed LLM training and inference are traditionally optimized in isolation; expert-crafted systems such as DeepEP, FLUX, and TokenWeave show the potential of co-design but require deep systems expertise and hardware-specific tuning; CUCo is an agentic framework that automates compute-communication co-design of CUDA kernels by combining a structured design-space formalization with a correctness-first fast-path agent for reliable baselines and an evolution-driven slow-path agent for high-performance strategies, achieving up to 1.

Tile-Level Activation Overlap for Efficient LLM Inference

arXiv:2607. 02521v1 Announce Type: cross Abstract: SwiGLU is the dominant MLP activation in modern large language models, yet its intermediate tensor materialization costs 9-37% of MLP execution time.