#model deployment
try —
3 papers match
cs.CE2026
Train Often, Deploy Selectively: Forward-Gated Model Replacement in Crypto Markets
Aditya Dutta
The paper proposes a deployment strategy called Shadow Before Swap (SBS) that evaluates retrained forecasting models off the live path and only promotes them when they show a suffi…
#model deployment#crypto markets#forecasting#online learning
cs.LG2026
Lottery Tickets Are Not Deployment Tickets
Bum Jun Kim
The paper investigates whether sparsified models (lottery tickets) can replace dense models in real deployment without altering downstream decision logic, finding that even with ma…
#sparsification#lottery tickets#model deployment#calibration
cs.AI2026
Cost-Optimal Foundation Model Deployment Portfolio for Transportation Management
Xi Cheng, Ke Liu, Siyuan Feng +2
The paper formulates and solves a cost‑optimal deployment problem for foundation models in transportation management centers, balancing quality, latency, safety, and GPU budget con…
#foundation models#transportation management#model deployment#cost optimization