AI Paper Atlas

Research map / Vision-language and multimodal / CLIP and zero-shot

Efficient and Versatile Robust Fine-Tuning of Zero-shot Models

Sungyeon Kim, Boseung Jeong, Donghyun Kim, Suha Kwak

ECCV 2024 · Vision-language and multimodal › CLIP and zero-shot · approach: Parameter-efficient visual tuning · 6 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
add adapter dropping to R-Adapter (ablation)R-Adapter · ImageNet ViT-B/32OOD avg accuracy: 47.7 -> 49.6Improved
add re-scaling (weight-space ensemble) to R-Adapter (ablation)R-Adapter · ImageNet ViT-B/32OOD avg accuracy: 47.7 -> 53.5 (ID 77.5 -> 76.5)Improved
add accumulation (EMA of adapter weights) (ablation)R-Adapter · ImageNet ViT-B/32OOD avg accuracy: 47.7 -> 47.8No clear effect
replace dropout or drop-path with adapter dropping (ablation)R-Adapter · ImageNet ViT-B/32OOD avg accuracy: DO 48.7, DP 47.9, AD 49.6Improved
replace InfoNCE with MPM-NCE (ablation)R-Adapter with AD+AC+RS · ImageNet ViT-B/32ID and OOD avg accuracy: ID 76.6 -> 77.5, OOD 53.7 -> 53.9Improved
add label smoothing to InfoNCE (ablation)R-Adapter · ImageNet ViT-B/32ID accuracy: 76.6 -> 76.9 ID is E7 vs E8; OOD 53.7 -> 54.0Hurt
add label smoothing to MPM-NCE (ablation)R-Adapter · ImageNet ViT-B/32ID and OOD avg accuracy: 77.5 -> 77.7 ID, 53.9 -> 54.3 OODImproved
use margin with single positive (no soft labels) (ablation)MPM-NCE · ImageNet ViT-B/32OOD avg accuracy: 54.3 -> 47.0Hurt
increase adapter rank from 4 to full (ablation)R-Adapter · ImageNet ViT-B/32ID / OOD avg accuracy: ID 72.5 -> 77.7, OOD 51.7 -> 54.3Improved
use adapter drop probability p=0.2 versus 0 (ablation)R-Adapter · ImageNet ViT-B/32OOD avg accuracy: 53.3 -> 54.3Improved
use R-Adapter instead of full fine-tuning or robust baselines with far fewer trainable parametersCLIP ViT-B/16 · ImageNet OOD avgtop-1 accuracy: Mask-Fill 63.3 -> 64.8Improved
use R-Adapter for open-vocab segmentation instead of OVSeg two-stage fine-tuningOVSeg · A-150mIoU: 29.6 -> 29.5No clear effect

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