Rethinking the Training Paradigm of Discrete Token-Based Multimodal LLMs: An Analysis of Text-Centric Bias.
Wansik Jo, Jooyeong Na, Soyeon Hong, Seungtaek Choi, Hyunsouk Cho
Browse the full CIKM paper archive.
Wansik Jo, Jooyeong Na, Soyeon Hong, Seungtaek Choi, Hyunsouk Cho
Browse the full CIKM paper archive.