[TMLR'26] Our first systematic survey on multimodal long-context token compression has been accepted to TMLR! Congrats to Kele, Keda, and all collaborators! [arxiv] [code]