CrossPool: Efficient Multi-LLM Serving for Cold MoE Models through KV-Cache and Weight Disaggregation
## Episode Summary In this episode, we cover: - **CrossPool: Efficient Multi-LLM Serving for Cold MoE Models through KV-Cache and Weight Disaggregation** (arXiv) - **Phase Matters: Characterizing Heterogeneous Vision-Language Inference on a Mobile SoC** (arXiv) - **AI is stress-testing processor architectures and RISC-V fits the moment - EDN - Voice of the Engineer** (google_riscv) - **Milk-V Jupiter2 is mini PC with a SpacemiT K3 RISC-V processor - Liliputing** (google_riscv) - **SpacemiT to launch server-class RISC-V processor following capital injection - South China Morning Post**…
The skinny
The skinny isn't ready yet — notes appear once the transcript is processed.