MMEdge: Accelerating On-device Multimodal Inference via Pipelined Sensing and Encoding Paper • 2510.25327 • Published Oct 29 • 1
MMEdge: Accelerating On-device Multimodal Inference via Pipelined Sensing and Encoding Paper • 2510.25327 • Published Oct 29 • 1 • 1
Which Heads Matter for Reasoning? RL-Guided KV Cache Compression Paper • 2510.08525 • Published Oct 9 • 22