Research map / Vision-language and multimodal / CLIP and zero-shot
Semantic Robustness Certification for Vision-Language Models
Peiyu Yang, Paul Montague, Feng Liu, Andrew Craig Cullen, Amardeep Kaur, Christopher Leckie, Sarah Monazam Erfani
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 ExactLine linear interpolation with semantic-plane transformation (image-specified) | CLIP certification · StanfordCars real-world, Color | mean absolute discrepancy (%): 32.8% -> 6.2% | Improved |
| replace ExactLine with semantic-plane transformation (image-specified) | CLIP certification · Flowers102 synthetic, Shape (OOD) | mean absolute discrepancy (%): 19.3% -> 6.6% | Improved |
| use text-specified instead of image-specified target extent (ablation) | semantic transformation · UCF101 real-world, Viewpoint | mean absolute discrepancy (%): 4.0% -> 14.1% | Hurt |
| use text-specified instead of image-specified target extent (ablation) | semantic transformation · DTD real-world, Texture | mean absolute discrepancy (%): 3.9% -> 2.4% | Improved |
| use our transformation instead of ExactLine on Caltech101 style | CLIP certification · Caltech101 real-world, Style | mean absolute discrepancy (%): 10.5% -> 11.9% (T-Spec), 7.8% (I-Spec) | Mixed |
| use our transformation instead of ExactLine on FGVCAircraft background | CLIP certification · FGVCAircraft real-world, Background | mean absolute discrepancy (%): 13.0% -> 14.9% (T-Spec), 11.0% (I-Spec) | Mixed |
| use our transformation instead of ExactLine on Food101 color (OOD) | CLIP certification · Food101 synthetic, Color (OOD) | mean absolute discrepancy (%): 9.3% -> 9.6% (T-Spec), 6.8% (I-Spec) | Mixed |
| increase misalignment budget delta (ablation) | certificate bounds · ImageNet subset | stable coverage | Hurt |
| increase misalignment budget delta (ablation) | certificate bounds · ImageNet subset | conditional invariance | No clear effect |
Closest papers by meaning
- VLM-RobustBench: A Comprehensive Benchmark for Robustness of Vision-Language Models (ICML 2026 · similarity 0.91)
- MVI-Bench: A Comprehensive Benchmark for Evaluating Robustness to Misleading Visual Inputs in LVLMs (ICML 2026 · similarity 0.89)
- SPA-VL: A Comprehensive Safety Preference Alignment Dataset for Vision Language Models (CVPR 2025 · similarity 0.89)
- Synthesize Diagnose and Optimize: Towards Fine-Grained Vision-Language Understanding (CVPR 2024 · similarity 0.89)
- An Empirical Study Into What Matters for Calibrating Vision-Language Models (ICML 2024 · similarity 0.89)
- ETA: Evaluating Then Aligning Safety of Vision Language Models at Inference Time (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.
