Research map / Vision-language and multimodal / CLIP and zero-shot
Benchmarking Robustness of Adaptation Methods on Pre-trained Vision-Language Models
Shuo Chen, Jindong Gu, Zhen Han, Yunpu Ma, Philip Torr, Volker Tresp
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 changed | Where | Measured | Effect |
|---|---|---|---|
| replace full fine-tuning with a single adapter | CLIP-BART · VQAv2 text corruptions | RR (%): 73.65 -> 77.64 | Improved |
| replace full fine-tuning with a single adapter | CLIP-BART · VQAv2 image corruptions | RR (%): 84.86 -> 85.76 | Improved |
| replace full fine-tuning with a single adapter | CLIP-BART · GQA image corruptions | RR (%): 89.20 -> 82.49 | Hurt |
| replace full fine-tuning with a single adapter | CLIP-BART · GQA clean | accuracy (%): 55.04 -> 54.14 | Hurt |
| use half-shared adapters instead of a single adapter | CLIP-BART · GQA image corruptions | RR (%): 82.49 -> 89.37 | Improved |
| use multiple adapters instead of a single adapter | CLIP-BART · GQA image corruptions | RR (%): 82.49 -> 86.16 | Improved |
| use multiple LoRA instead of single LoRA | CLIP-BART · GQA image corruptions | RR (%): 82.58 -> 91.15 | Improved |
| use multiple LoRA instead of single LoRA | CLIP-BART · GQA text corruptions | RR (%): 63.94 -> 68.77 | Improved |
| use Hyperformer instead of full fine-tuning | CLIP-BART · GQA text corruptions | RR (%): 66.92 -> 70.30 | Improved |
| apply text corruptions instead of image corruptions | Single Adapter on CLIP-BART · VQAv2 | RR (%): 85.76 -> 77.64 | Hurt |
| increase adaptation data from 20% to 100% (ablation) | adaptation methods · all tasks | relative robustness | Hurt |
| increase soft prompt length (ablation) | prompt tuning on CLIP-BART · four tasks | clean performance | Improved |
Closest papers by meaning
- VLM-RobustBench: A Comprehensive Benchmark for Robustness of Vision-Language Models (ICML 2026 · similarity 0.90)
- The Illusion of Progress? A Critical Look at Test-Time Adaptation for Vision-Language Models (NeurIPS 2025 · similarity 0.90)
- Realistic Test-Time Adaptation of Vision-Language Models (CVPR 2025 · similarity 0.90)
- Machine Vision Therapy: Multimodal Large Language Models Can Enhance Visual Robustness via Denoising In-Context Learning (ICML 2024 · similarity 0.89)
- Scaling Test-Time Robustness of Vision-Language Models via Self-Critical Inference Framework (CVPR 2026 · similarity 0.89)
- Adapting Multi-modal Large Language Model to Concept Drift From Pre-training Onwards (ICLR 2025 · similarity 0.88)
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.
