New KV cache compaction technique cuts LLM memory 50x without accuracy loss

MIT researchers developed Attention Matching, a KV cache compaction technique that compresses LLM memory by 50x in seconds — ...
Nature

A region-matching method for pulse transit time estimation: potential for improving the accuracy in determining carotid femoral pulse wave velocity

Using a region of pulse waveform to estimate PTT may increase the accuracy in cfPWV measurement. A region-based method that obtained PTT by shifting the proximal pulse wave and superimposing its ...
HUB

Groundbreaking method to match celestial objects across telescopes

A team of Johns Hopkins researchers has developed a cutting-edge data science approach capable of matching observations of celestial objects taken across multiple telescope surveys, overcoming a ...