Tag
论文提出 LSA-PAST，一种面向组合式增量学习（composition-IL）的参数高效方法，通过在冻结 ViT 中注入深度特化的 LoRA 适配器并用无复习的仿射变换将旧原型映射到新适配器空间，实现了精度-遗忘权衡的 state-of-the-art 表现。
CoDrift is a compositional framework for offline reinforcement learning that combines behavioral and value objectives into a unified policy field, achieving state-of-the-art performance on benchmarks like OGBench and D4RL.
Shruti Joshi invites attendees to a poster session at UncertaintyInAI to discuss research on improving Sparse Autoencoder (SAE) architectures for better out-of-distribution generalization.