HyperAdapter performs PEFT of ViTs via soft hypergraph construction, hyperedge-level bottleneck adaptation, and incidence-based diffusion, claiming consistent gains over token-wise adapters on structured visual benchmarks.
Efficient learn- ing with sine-activated low-rank matrices
4 Pith papers cite this work. Polarity classification is still indexing.
4
Pith papers citing it
years
2026 4representative citing papers
Replacing the linear Query projection with an identity-plus-bottleneck-MLP residual improves GPT-3-small-style validation log-loss by 2.40% and perplexity by 6.81%.
A plug-and-play KL regularizer that masks the target token and renormalizes probabilities to improve the learning-forgetting trade-off in LoRA adaptation of LLMs.
SegTTA improves MedSAM2 zero-shot segmentation on uterus and liver datasets by test-time augmentations plus weighted voting, delivering +1.6 mIoU and -2.0 HD95 on multiclass hepatic vessels.
citing papers explorer
-
Structured Hyperedge Adaptation for Parameter-Efficient Fine-Tuning of Vision Transformers
HyperAdapter performs PEFT of ViTs via soft hypergraph construction, hyperedge-level bottleneck adaptation, and incidence-based diffusion, claiming consistent gains over token-wise adapters on structured visual benchmarks.
-
Beyond Linearity in Attention Projections: The Case for Nonlinear Queries
Replacing the linear Query projection with an identity-plus-bottleneck-MLP residual improves GPT-3-small-style validation log-loss by 2.40% and perplexity by 6.81%.
-
Mask the Target: A Plug-and-Play Regularizer Against LoRA Forgetting
A plug-and-play KL regularizer that masks the target token and renormalizes probabilities to improve the learning-forgetting trade-off in LoRA adaptation of LLMs.
-
SegTTA: Training-Free Test-Time Augmentation for Zero-Shot Medical Imaging Segmentation
SegTTA improves MedSAM2 zero-shot segmentation on uterus and liver datasets by test-time augmentations plus weighted voting, delivering +1.6 mIoU and -2.0 HD95 on multiclass hepatic vessels.