On the Design Fundamentals of Pixel Text Representation Learning
Paper • 2609.01147 • Published • 29
Pixel Linguist II
Vision-only encoders for language representation directly in pixel space.
-
Pixel-Linguist/Pixel-Linguist-II-Pretrain
Feature Extraction • 0.7B • Updated • 45 • 1Note Stage 1: foundational multilingual pretraining.
Pixel-Linguist/Pixel-Linguist-II-Midtrain
Feature Extraction • 0.7B • Updated • 81 • 2Note Stage 1 + Stage 2: full curriculum.
Pixel-Linguist/Pixel-Linguist-II-Midtrain-Only
Feature Extraction • 0.7B • Updated • 91 • 3Note Stage 2 only: paper headline English STS and ViDoRe model.