In specific TinyML applications, researchers have deployed heavily quantized models, such as DeepConv LSTMs, onto resource-constrained microcontrollers (e.g., Arduino Nano 33 BLE Sense), achieving ...