GRASP applies deterministic conditioning-space partitioning and sample-wise residual adapters to improve tail-class fidelity, diversity, and downstream utility in flow matching models, outperforming full fine-tuning and MoE baselines on medical and ImageNet long-tail data.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2025 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
GRASP: Guided Residual Adapters with Sample-wise Partitioning
GRASP applies deterministic conditioning-space partitioning and sample-wise residual adapters to improve tail-class fidelity, diversity, and downstream utility in flow matching models, outperforming full fine-tuning and MoE baselines on medical and ImageNet long-tail data.