TMLR

PruneFuse: Efficient Data Selection via Weight Pruning and Network Fusion

PruneFuse：通过权重剪枝和网络融合实现高效的数据选择

Humaira Kousar, Hasnain Irshad Bhatti, Jaekyun Moon

AI总结 PruneFuse通过剪枝网络进行数据选择并融合原始网络，降低计算成本，提升训练效率，优于基线方法。

Comments Published in TMLR (Featured Certification). arXiv admin note: substantial text overlap with arXiv:2501.01118

Journal ref Transactions on Machine Learning Research (TMLR), March 2026

Do All Vision Transformers Need Registers? A Cross-Architectural Reassessment

视觉变换器是否都需要寄存器？一种跨架构重新评估

Spiros Baxevanakis, Platon Karageorgis, Ioannis Dravilas, Konrad Szewczyk

AI总结 本文重新评估了视觉变换器中寄存器的必要性，验证了Darcet等人的发现，并发现其结论在不同模型中并不普遍，同时探讨了模型规模和术语一致性的影响。

Comments Preprint. Submitted to Transactions on Machine Learning Research (TMLR). 26 pages, 17 figures