MeshKV: A Network-on-Chip KV Cache Fabric for Scalable Transformer Decoding Accelerators
This paper presents a hardware architecture for scaling transformer inference, a capability advance with no direct safety or governance implications.
This paper presents a hardware architecture for scaling transformer inference, a capability advance with no direct safety or governance implications.
Auto-scored: Technical accelerator paper representing incremental capability progress without misuse or governance relevance.