Research map / Vision-language and multimodal / CLIP and zero-shot
Post-hoc Probabilistic Vision-Language Models
Anton Baumann, Rui Li, Marcus Klasson, Santeri Mentu, Shyamgopal Karthik, Zeynep Akata, Arno Solin, Martin Trapp
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 |
|---|---|---|---|
| apply BayesVLM post-hoc Laplace uncertainty to CLIP | OpenCLIP ViT-B-32 · CIFAR-100 zero-shot | ECE (%): 6.31 -> 1.92 | Improved |
| apply BayesVLM post-hoc Laplace uncertainty to CLIP | OpenCLIP ViT-B-32 · UCF101 zero-shot | ECE (%): 11.52 -> 3.57 | Improved |
| apply BayesVLM post-hoc Laplace uncertainty to CLIP | OpenCLIP ViT-B-32 · Flowers-102 zero-shot | ECE (%): 6.59 -> 4.22 | Improved |
| apply BayesVLM post-hoc Laplace uncertainty to CLIP | OpenCLIP ViT-B-32 · Flowers-102 zero-shot | accuracy (%): 68.99 -> 68.87 | No clear effect |
| apply BayesVLM post-hoc Laplace uncertainty to CLIP | OpenCLIP ViT-B-32 · UCF101 zero-shot | NLPD: 1.59 -> 1.44 | Improved |
| use BayesVLM instead of test-time augmentation (TTA) | OpenCLIP ViT-B-32 · CIFAR-10 zero-shot | accuracy (%): 88.54 -> 93.62 | Hurt |
| add ProbCosine to PCME++ probabilistic embeddings | PCME++ · Flowers-102 zero-shot | ECE (%): 8.81 -> 2.79 | Improved |
| add ProbCosine to PCME++ probabilistic embeddings | PCME++ · SUN397 zero-shot | ECE (%): 26.03 -> 19.41 | Improved |
| add ProbCosine to PCME++ probabilistic embeddings | PCME++ · Flowers-102 zero-shot | accuracy (%): 40.59 -> 40.43 | No clear effect |
| estimate the Hessian on proxy dataset CC12M instead of LAION-400M (ablation) | BayesVLM · Flowers-102 zero-shot | accuracy (%): 68.87 -> 68.12 | No clear effect |
| estimate the Hessian on proxy dataset CC12M instead of LAION-400M (ablation) | BayesVLM · SUN397 zero-shot | ECE (%): 2.06 -> 3.89 | Mixed |
| use BayesVLM uncertainties (EPIG, BALD) for support-set selection instead of random or entropy | CLIP fine-tuning · OfficeHome and ImageNet-variant active learning | weighted accuracy and NLPD | Improved |
Closest papers by meaning
- Exploiting the Asymmetric Uncertainty Structure of Pre-trained VLMs on the Unit Hypersphere (NeurIPS 2025 · similarity 0.92)
- Probabilistic Language-Image Pre-Training (ICLR 2025 · similarity 0.92)
- Revisiting the Role of Language Priors in Vision-Language Models (ICML 2024 · similarity 0.89)
- An Empirical Study Into What Matters for Calibrating Vision-Language Models (ICML 2024 · similarity 0.89)
- FUSE: Quantifying Uncertainty in Vision-Language Models by Bayesian Fusing Epistemic and Aleatoric Uncertainty (ICML 2026 · similarity 0.89)
- Fine-Grained Post-Training Quantization for Large Vision Language Models with Quantization-Aware Integrated Gradients (CVPR 2026 · 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.
