Recent research from Arm
Explore our latest research publications and technical explorations.
LFM-LLIE for low-light image enhancement
Explore methods for improving image clarity in difficult lighting environments.
LATMiX for microscaling quantization
See how to improve post-training quantization for more efficient LLM deployment.
HeatKV for visual autoregressive modeling
Discover approaches that reduce KV-cache demands in image generation.
Attention drift in speculative decoding
See how normalization improves speculative decoding robustness.
Workload churn in design space exploration
Learn how workload churn can shape long-term performance in SoC architecture.
Solving cross-NUMA performance issues
See how cross-NUMA optimizations accelerate LLM inference on Arm Neoverse.
Rethinking robotics reinforcement learning
Explore a practical workflow for humanoid training and deployment.
Optimizing vLLM inference on CPU
Learn how to optimize a high-throughput inference engine on Arm CPUs.
A caution about trust in a world of agentic AI
Why we must not trivialize the word “trust”, and how agentic AI may be a blessing for trust.