Running 28 Weight-Space Geometry of Offline Reasoning Training 🧭 28 Interactive weight-space geometry of six reasoning losses
777Radik/potion-multilingual-128M-int8 Sentence Similarity • 64M • Updated about 1 month ago • 13 • 1
777Radik/potion-multilingual-128M-int8 Sentence Similarity • 64M • Updated about 1 month ago • 13 • 1
AudioSAE: Towards Understanding of Audio-Processing Models with Sparse AutoEncoders Paper • 2602.05027 • Published Feb 4 • 63
Unveiling Intrinsic Dimension of Texts: from Academic Abstract to Creative Story Paper • 2511.15210 • Published Nov 19, 2025 • 91