1 paper
Jiawei Guo, Yu Chen, Xiang Wang +6
Recent latent visual reasoning methods achieve substantial gains by inserting continuous latent tokens into multimodal language models. These gains are commonly attributed to the t…