AI Paper Atlas

Research map / Federated learning and privacy / Personalized and heterogeneous FL

Latte: Collaborative Test-Time Adaptation of Vision-Language Models in Federated Learning

Wenxuan Bao, Ruxi Deng, Ruizhong Qiu, Tianxin Wei, Hanghang Tong, Jingrui He

ICCV 2025 · Federated learning and privacy › Personalized and heterogeneous FL · approach: Federated prompt and LoRA tuning · 5 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
share one global memory across all clientsTDA · VLCS (ViT-B/16)accuracy (%): 81.44 -> 80.29Hurt
share one global memory across all clientsDMN-ZS · VLCS (ViT-B/16)accuracy (%): 81.12 -> 80.55Hurt
share one global memory across all clientsDMN-ZS · CIFAR-10-C (ViT-B/16)accuracy (%): 67.42 -> 63.90Hurt
share one global memory across all clientsDMN-ZS · TerraIncognita (ViT-B/16)accuracy (%): 33.65 -> 37.64Improved
share one global memory across all clientsTDA · TerraIncognita (RN50)accuracy (%): 36.73 -> 37.78Improved
use Latte local plus external memory instead of global sharingDMN-ZS (global) · TerraIncognita (ViT-B/16)accuracy (%): 37.64 -> 40.95Improved
use Latte instead of best baseline ZeroZero · VLCS (ViT-B/16)accuracy (%): 81.98 -> 82.57Improved
use Latte instead of CLIP zero-shotCLIP · TerraIncognita (RN50)accuracy (%): 23.24 -> 41.47Improved
use Latte instead of CLIP zero-shotCLIP · CIFAR-100-C (ViT-B/16)accuracy (%): 38.91 -> 39.51Improved
add memory-based TTA on CIFAR-100-C vs episodic TPT/ZeroZero · CIFAR-100-C (ViT-B/16)accuracy (%): 38.91 -> 37.55Hurt
update external memory every T samples up to 50 (ablation)Latte · TerraIncognitaaccuracy (%): T<=50 unchanged; T = 200 slight dropNo clear effect
add external memory to local memory (ablation)Latte (local only) · VLCSaccuracy (%): local only or external only -> merged (no numbers given)Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy