Early-Signal Detector

method · first seen 4 Oct 2026

visual token reduction

Also: token compression, token condensation, visual token pruning, visual token routing, spatial token pruning, image-token compression, pixel-space compression

Pruning or merging image tokens in vision-language models to cut inference cost while preserving accuracy.

Mentions per week

02119 May11 Aug10 NovPapersCodeDiscussion

PapersCodeDiscussion

Rank 28 on 16 Nov 2026 · 30 mentions in 28 days · 30 independent groups · papers · first seen 6 weeks ago · acceleration +1.40

Earliest sources

Latest sources

Mentioned together with

vision-language-action modelvision-language modelmultimodal large language model physics reasoningKV cache reuseretrieval-augmented generationGUI agentknowledge distillationaction chunkingcertified accelerationjailbreak attacks on vision-language models