↓ Skip to main content

ML & AI

Measuring Thinking Efficiency in Reasoning Models: The Missing Benchmark ↗ ↖
ML & AI Artificial Intelligence LLMs
(Guest article on the Nous Research blog) Anecdotal evidence suggests open weight models produce significantly more tokens for similar tasks than closed weight models. This report systematically investigates these observations. We confirm this trend to be generally true, but observe significant differences depending on problem domain.
Neural Network Visualization
·148 words·1 min
ML & AI Machine Learning
A browser based interactive application that visualizes simple multi-layer perception (MLP) neural networks for the inference of 8x8 pixel images.
BitNetPDK: Neural Networks (MNIST inference) on the "3-cent" Microcontroller
·1227 words·6 mins
ML & AI Microcontroller Edge ML Machine Learning Padauk
Is it possible to implement reasonably accurate inference of MNIST, the handwritten numbers dataset, on a “3 cent” Microcontroller with only 64 bytes of RAM and 1K of instruction memory?
BitNetMCU: Implementing Neural Networks on the “10-cent” RISC-V MCU without Multiplier
·972 words·5 mins
ML & AI Microcontroller Edge ML Machine Learning RISC-V
BitNetMCU is a project focused on the training and inference of low-bit quantized neural networks, designed to run efficiently on low-end microcontrollers like the CH32V003. Quantization aware training (QAT) and fine-tuning of model structure allowed surpassing 99% Test accuracy on a 16x16 MNIST dataset in only 2kb of RAM and 16kb of Flash.