Research map / Vision-language and multimodal / CLIP and zero-shot
BayesAdapter: Enhanced Uncertainty Estimation in CLIP Few-Shot Adaptation
Pablo Morales-Álvarez, Stergios Christodoulidis, Maria Vakalopoulou, Pablo Piantanida, José Dolz
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 |
|---|---|---|---|
| use Bayesian inference over the parameters instead of MAP point estimate | CLAP · 11 datasets avg (ResNet50, all shots) | ECE: 6.592 -> 4.151 | Improved |
| use Bayesian inference instead of MAP | CLAP · 11 datasets avg (ViT-B/16, all shots) | ECE: 5.351 -> 3.835 | Improved |
| use Bayesian inference instead of MAP | CLAP · 11 datasets avg (ResNet50, all shots) | coverage at 99% confidence: 7.103 -> 13.423 | Improved |
| use Bayesian inference instead of MAP | CLAP · 11 datasets avg (ResNet50, all shots) | Accuracy: 70.182 -> 69.476 | Hurt |
| use Bayesian inference instead of MAP | CLAP · 11 datasets avg (ViT-B/16, all shots) | Accuracy: 76.980 -> 76.420 | Hurt |
| increase to 32 shots per class | BayesAdapter · 11 datasets (32 shots) | Accuracy | Improved |
| increase number of MC samples from 1 to 10 (ablation) | BayesAdapter · 11 datasets avg (ablation, ResNet50) | Accuracy: 69.443 -> 69.420 | No clear effect |
| increase prior std from 0.0001 to 0.01 (ablation) | BayesAdapter · 11 datasets avg (ablation, ResNet50) | Accuracy: 58.693 -> 69.437 | Improved |
| increase prior std from 0.01 to 1.0 (ablation) | BayesAdapter · 11 datasets avg (ablation, ResNet50) | Accuracy: 69.437 -> 68.747 | Hurt |
| increase prior std from 0.01 to 1.0 (ablation) | BayesAdapter · 11 datasets avg (ablation, ResNet50) | Cov@99% confidence: 9.213 -> 12.043 | Improved |
| add Bayesian variance parameters (training cost) | CLAP · Caltech101 training time, 32 shots | training time (s): 32.25 -> 42.40 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- CLAP — CLAP: Isolating Content from Style through Contrastive Learning with Augmented Prompts (ECCV 2024) · better on 6 of 8 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Bayesian Exploration of Pre-trained Models for Low-shot Image Classification (CVPR 2024 · similarity 0.91)
- A Hard-to-Beat Baseline for Training-free CLIP-based Adaptation (ICLR 2024 · similarity 0.90)
- Bayesian Test-Time Adaptation for Vision-Language Models (CVPR 2025 · similarity 0.90)
- What Drives Test-Time Adaptation for CLIP? A Controlled Empirical Study from an Update Perspective (NeurIPS 2026 · similarity 0.89)
- Adaptive Parameter Selection for Tuning Vision-Language Models (CVPR 2025 · similarity 0.89)
- CLIPTTA: Robust Contrastive Vision-Language Test-Time Adaptation (NeurIPS 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.
