In federated remote sensing, LoRA tuning of a frozen CLIP model achieves the best accuracy-to-communication trade-off, while full fine-tuning causes severe catastrophic forgetting of pretrained knowledge.
Learning transferable visual models from natural language supervision,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
On the Effectiveness of Adaptation Strategies for VLM-Based Federated Learning in Remote Sensing
In federated remote sensing, LoRA tuning of a frozen CLIP model achieves the best accuracy-to-communication trade-off, while full fine-tuning causes severe catastrophic forgetting of pretrained knowledge.