THE BENCH · webgpu-kv-thermal
WebGPU KV-Cache Thermal Profiler
Time to first token as quadratic causal attention saturates the local WebGPU adapter, then under AST DAG context pruning.
Interactive Lab · webgpu-kv-thermalWebGPU KV-Cache Attention Thermal Profiler
Probe hardware WebGPU adapters, visualize O(N²) causal attention thermal saturation, and run live AST DAG context pruning to snap TTFT back under 200ms.