@HuggingPapers: YOLO-PEFT Tencent researchers propose a structure-aware framework that turns adapter placement into auditable constrain…

X AI KOLs Timeline Papers

Summary

Tencent researchers propose YOLO-PEFT, a structure-aware framework that treats adapter placement as constraint planning, outperforming full fine-tuning on YOLO detectors while using significantly fewer trainable parameters.

YOLO-PEFT Tencent researchers propose a structure-aware framework that turns adapter placement into auditable constraint planning, beating full fine-tuning on YOLO detectors while using far fewer trainable parameters. https://t.co/o8vZ8C9CZa
Original Article
View Cached Full Text

Cached at: 08/10/26, 11:37 PM

YOLO-PEFT

Tencent researchers propose a structure-aware framework that turns adapter placement into auditable constraint planning, beating full fine-tuning on YOLO detectors while using far fewer trainable parameters. https://t.co/o8vZ8C9CZa

Similar Articles

YOLO-PEFT: Parameter-Efficient Fine-Tuning on YOLO Family

Hugging Face Daily Papers

YOLO-PEFT is a structure-aware framework that formulates adapter placement as constraint planning for parameter-efficient fine-tuning of YOLO detectors, achieving better mAP than full fine-tuning with reduced memory.

PFAdapter: Hierarchical LoRA Decomposition for Personalized Federated MLLMs

arXiv cs.LG

This paper introduces PFAdapter, a communication-efficient framework for personalized federated fine-tuning of Multimodal Large Language Models (MLLMs). It uses hierarchical LoRA decomposition to separate adapter parameters into global-shared and local-private components, achieving near 50% reduction in communication costs while improving personalization through orthogonality regularization.

PorTAL: Portable Task Adapters for LLMs (3 minute read)

TLDR AI

PorTAL is a novel architecture that decouples task fine-tuning from specific base model weights, enabling portable task adapters that can be transferred to new models with minimal retraining. It achieves ~98% of LoRA's accuracy gain on unseen models using only half the calibration data.