Machine Learning Blog

Explore how Arm adapters for Google Model Explorer help developers visualize AI models, inspect graph transformations, and identify performance bottlenecks

Explore how CMSIS-Ethos-U, the SDS Framework, and CMSIS Developer Assistant simplify Edge AI deployment, data replay, integration and debugging on devices

Explore how ExecuTorch, Zephyr and Arm Ethos-U55 bring PyTorch models to life on the Renesas RA8P1 through a practical camera-based edge AI application

Learn how Arm and Google optimize ML Drift for Arm GPUs, improving on-device AI inference performance through kernel, resource, and workgroup tuning

See how LiteRT, quantization, KleidiAI, and SME2 improve on-device toxic speech detection on mobile, reducing latency and server processing demand overall

Explore how Llama-Mobile uses S3D8 and quantization-aware training to reduce model size while retaining accuracy for multimodal AI on Arm mobile CPUs

How the on-device Smart Gallery enables more natural, private user experiences on Android.

Discover how Arm researchers used generative AI to improve low-light image enhancement for real-world AI vision systems.

End-to-end guide to training, quantizing, and deploying an int8 Conformer speech model on Arm, using ExecuTorch and Ethos-U85 for efficient edge AI today.

A detailed look at how CPU-based embedding, unified memory, and local retrieval workflows come together to enable responsive, private RAG pipelines on desktop AI platforms—powered by Arm and DGX Spark…

Part 4 of the series describes the new SME2 multi-vector and Lookup table features and instructions

Discover how Arm’s Model Training Gym helps you train, fine-tune, and export neural graphics models for real-time rendering with PyTorch and ExecuTorch.

LLVM 21.1.0 was released on August 26, 2025 with contributions from Arm engineers to improve performance. This post provides the key highlights.

In this blog post, learn how the Arm Developer Advocacy team is exploring how machine learning can enable a sign-to-speech translator for video conferencing.

A deep dive into a practical, ML-powered approach to temporal super sampling.

Build next-gen mobile AI apps with SME2—no code changes needed. Accelerate performance across devices using top AI frameworks and runtimes.

A year of Arm KleidiAI in XNNPack brings major ML performance boosts—no code changes needed. Transparent, seamless acceleration on Arm CPUs.

In this blog post, we perform a study of the GNN architecture and the new TF-GNN API and determine whether GNNs are a viable approach for implementing physics simulations.