ByteDance-Seed/UI-TARS-7B-SFT
Image-Text-to-Text • 8B • Updated • 839 • 181
None defined yet.
SAE Interprets MoE: Sparse AutoEncoders Reveal That MoE Experts Exhibit Human-Interpretable Fine-Grained Specialization
Your LR Schedule May Not Be Principled: Effective Learning Rate Scheduling in Weight-Normalized Training