Ultra-Quantisation: Efficient Embedding Search via 1.58-bit Encodings

MQRe-assign community