Tagged “model deployment”
Research
Can Pruning Mixture-of-Experts Models Reduce Deployment Memory?
A new paper reports that pruned MoE models hold up on benchmarks at 75% and 50% expert retention, but it does not measure memory, latency, or throughput.
Research
AgentDP’s “Up to 80% Faster” Claim Needs a Hardware Matrix
The abstract reports up to 80% less adaptation time and 13%+ higher accuracy, but not which devices, models, or calculations sit behind those numbers.