Academic

Academic

Academic · 약 1분

Improving Robustness In Sparse Autoencoders via Masked Regularization

arXiv:2604.06495v1 Announce Type: new Abstract: Sparse autoencoders (SAEs) are widely used in mechanistic interpretability to project LLM activations onto sparse latent spaces. However, sparsity alone …

Vivek Narayanaswamy, Kowshik Thopalli, Bhavya Kailkhura, Wesam Sakla
조회수 116회
Academic · 약 1분

Faster Superword Tokenization

arXiv:2604.05192v1 Announce Type: new Abstract: Byte Pair Encoding (BPE) is a widely used tokenization algorithm, whose tokens cannot extend across pre-tokenization boundaries, functionally limiting it …

Craig W. Schmidt, Chris Tanner, Yuval Pinter
조회수 63회
Academic · 약 1분

PRIME: Prototype-Driven Multimodal Pretraining for Cancer Prognosis with Missing Modalities

arXiv:2604.04999v1 Announce Type: new Abstract: Multimodal self-supervised pretraining offers a promising route to cancer prognosis by integrating histopathology whole-slide images, gene expression, and pathology reports, …

Kai Yu, Shuang Zhou, Yiran Song, Zaifu Zhan, Jie Peng, Kaixiong Zhou, Tianlong Chen, Feng Xie, Meng Wang, Huazhu Fu, Mingquan Lin, Rui Zhang
조회수 59회