AI RESEARCH
Tempora: Characterising the Time-Contingent Utility of Online Test-Time Adaptation
arXiv CS.LG
•
ArXi:2602.06136v2 Announce Type: replace Test-time adaptation (TTA) offers a compelling remedy for machine learning (ML) models that degrade under domain shifts, improving generalisation on-the-fly with only unlabelled samples. This flexibility suits real deployments, yet conventional evaluations unrealistically assume unbounded processing time, overlooking the accuracy-latency trade-off. As ML increasingly underpins latency-sensitive and user-facing use-cases, temporal pressure constrains the viability of adaptable inference; predictions arriving too late to act on are futile. We.