Early-Signal Detector

method · first seen 5 Oct 2026

token pruning

Also: ToMe, patch pruning, token merging, token reduction, token sparsification, bipartite token merging

Removing less relevant tokens during model inference to reduce computation with limited loss in performance.

Mentions per week

0919 May11 Aug10 NovPapersCodeDiscussion

PapersCodeDiscussion

Earliest sources

Latest sources

Mentioned together with

vision transformertraining-free model accelerationknowledge distillationdistribution shiftmultimodal large language model physics reasoningatmospheric foundation modelKV cache evictionvisual token reductionworld action modeltest-time adaptation