Product Information
Description
An annual survey of the cs.PF arXiv category, covering research published from September 2025 – August 2026. Built on a close-reading set of 120 papers and organised into 44 chapters, it maps the year by the field's research sub-directions.
Book framework (share of the year's output):
- LLM Inference Optimization and Serving Systems (14.8%) — Focuses on optimizing large language model inference latency, throughput, and scheduling through techniques like compression, ranking, and specialized serving architectures.
- Edge AI and Distributed Resource Management (14.8%) — Addresses latency, resource allocation, and service orchestration challenges in edge and cloud environments for AI workloads.
- High-Performance Computing and Code Optimization (13.9%) — Covers performance analysis, auto-tuning, and optimization of high-performance computing kernels and code frameworks.
- Sparse Linear Algebra and Matrix Multiplication (11.3%) — Explores efficient algorithms and hardware acceleration for sparse matrix operations, tensor decompositions, and low-rank approximations.
- GPU Memory Hierarchy and Bandwidth Optimization (10.4%) — Investigates memory access patterns, bandwidth regulation, and CPU-GPU co-execution strategies to mitigate memory bottlenecks in heterogeneous systems.
- KV Cache Management for Long-Context LLMs (9.5%) — Develops techniques for efficient key-value cache management and attention optimization to support long-context inference in transformer models.
- Energy Efficiency and Power Modeling in AI (7.6%) — Analyzes energy consumption patterns, power modeling methodologies, and energy-efficient hardware/software co-design for AI workloads.
- Data Compression and Scalable Learning (6.7%) — Examines lossy compression techniques, error propagation, and scalable algorithms for data mining and machine learning at large scales.
- Job Scheduling and Queueing Theory (6.5%) — Applies queueing theory and scheduling policies to optimize job execution times and resource utilization in server and cloud environments.
- Quantum Computing and Post-Quantum Security (4.6%) — Benchmarks quantum algorithms, post-quantum cryptography performance, and hybrid quantum-classical computing architectures.
Each sub-direction is covered across three chapters — how it framed its questions, the strength of its key findings, and its open debates and frontiers — so the reader gets a traceable picture of what this part of the field actually produced over the year.
