Research map / Label-efficient and robust learning / Recognition and long-tail
SpotTune: Transfer Learning Through Adaptive Fine-Tuning
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 changed | Where | Measured | Effect |
|---|---|---|---|
| use input-dependent freeze/fine-tune routing instead of standard fine-tuning | ImageNet-pretrained ResNet-50 · CUBS | accuracy: 81.86% -> 84.03 % | Helped |
| use input-dependent freeze/fine-tune routing instead of standard fine-tuning | ImageNet-pretrained ResNet-50 · Stanford Cars | accuracy: 89.74% -> 92.40% | Helped |
| use input-dependent freeze/fine-tune routing instead of standard fine-tuning | ImageNet-pretrained ResNet-50 · Flowers | accuracy: 93.67% -> 96.34% | Helped |
| use input-dependent freeze/fine-tune routing instead of standard fine-tuning | ImageNet-pretrained ResNet-50 · WikiArt | accuracy: 75.60% -> 75.77% | Mixed |
| use input-dependent freeze/fine-tune routing instead of standard fine-tuning | ImageNet-pretrained ResNet-50 · Sketches | accuracy: 79.58% -> 80.20% | Helped |
| route all test images through fine-tuned blocks rather than selecting pretrained or fine-tuned blocks per image (ablation) | SpotTune · CUBS | accuracy: 84.03 % -> 82.36% | Hurt |
| route all test images through fine-tuned blocks rather than selecting pretrained or fine-tuned blocks per image (ablation) | SpotTune · Stanford Cars | accuracy: 92.40% -> 92.04% | Hurt |
| route all test images through fine-tuned blocks rather than selecting pretrained or fine-tuned blocks per image (ablation) | SpotTune · Flowers | accuracy: 96.34% -> 93.49% | Hurt |
| replace image-specific routing with a shared learned policy fine-tuning k blocks (ablation) | SpotTune · Visual Decathlon | score: 3612 -> 3401 | Hurt |
| replace image-specific routing with a shared learned policy fine-tuning k blocks (ablation) | SpotTune · Flowers | accuracy: 96.34% -> 96.60% | Helped |
| replace image-specific routing with a shared learned policy fine-tuning k blocks (ablation) | SpotTune · Sketches | accuracy: 80.20% -> 80.02% | Hurt |
Closest papers by meaning
- Facing the Elephant in the Room: Visual Prompt Tuning or Full finetuning? (ICLR 2024 · similarity 0.83)
- LoKi: Low-dimensional KAN for Efficient Fine-tuning Image Models (CVPR 2025 · similarity 0.83)
- LEAD: Exploring Logit Space Evolution for Model Selection (CVPR 2024 · similarity 0.83)
- RandLoRA: Full rank parameter-efficient fine-tuning of large models (ICLR 2025 · similarity 0.83)
- Quick-Tune: Quickly Learning Which Pretrained Model to Finetune and How (ICLR 2024 · similarity 0.83)
- Skip Tuning: Pre-trained Vision-Language Models are Effective and Efficient Adapters Themselves (CVPR 2025 · similarity 0.83)
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.