Different valid temporal partitions of the same streaming dataset can produce materially different rankings and performance numbers for continual learning methods.
Proceedings of the Thirty-Second AAAI Conference on Artificial Intelligence (AAAI) , year =
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Temporal Taskification in Streaming Continual Learning: A Source of Evaluation Instability
Different valid temporal partitions of the same streaming dataset can produce materially different rankings and performance numbers for continual learning methods.