Research map / Vision-language and multimodal / CLIP and zero-shot
Mitigating Endogenous Confirmation Bias in Noisy Label Learning for Vision-Language Models
Feiyang Ning, Xinyang Chen
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 |
|---|---|---|---|
| add Phase 2 noisy label refinement (ablation) | DKAF · Avg of CIFAR-100, Cars, CUB (Sym. 0.6) | accuracy: 82.57 -> 85.23 | Improved |
| add Phase 3 full fine-tuning on clean and corrected samples (ablation) | DKAF · Avg of CIFAR-100, Cars, CUB (Sym. 0.6) | accuracy: 75.38 -> 82.57 | Improved |
| add Phase 1 clean sample selection (ablation) | DKAF · Avg of CIFAR-100, Cars, CUB (Sym. 0.6) | accuracy: 69.39 -> 75.38 | Improved |
| add Phase 3 on Stanford-Cars (ablation) | DKAF · Stanford-Cars (Sym. 0.6) | accuracy: 77.04 -> 86.94 | Improved |
| add Phase 2 on CIFAR-100 (ablation) | DKAF · CIFAR-100 (Sym. 0.6) | accuracy: 86.41 -> 87.85 | Improved |
| use DKAF instead of second-best method | DKAF · CUB-200-2011 (Sym. 0.6) | accuracy: 73.08 -> 78.22 | Improved |
| use DKAF instead of second-best method | DKAF · OxfordPets (Sym. 0.6) | accuracy: 88.67 -> 92.72 | Improved |
| use DKAF instead of second-best method | DKAF · Stanford-Cars (Sym. 0.6) | accuracy: 86.35 -> 89.61 | Improved |
| use DKAF instead of DEFT on real-world noisy data | DKAF · CIFAR-100N | accuracy: 79.04 -> 81.91 | Improved |
| use DKAF instead of DEFT on real-world noisy data | DKAF · Clothing-1M | accuracy: 72.44 -> 73.36 | Improved |
| use DKAF instead of DEFT on real-world noisy data | DKAF · WebVision | accuracy: 85.12 -> 85.35 | No clear effect |
| use cross-modal collaborative pseudo-labeling instead of DEFT self-training on hard class Bombay | DKAF · OxfordPets Bombay class (40% noise) | F1-score: 15.87 -> 99.15 | Improved |
Closest papers by meaning
- Vision-Language Models are Strong Noisy Label Detectors (NeurIPS 2024 · similarity 0.90)
- Language Bias in LVLMs: From In-Depth Analysis to Simple and Effective Mitigation (ICML 2026 · similarity 0.89)
- Pre-Trained Vision-Language Models as Noisy Partial Annotators (AAAI 2025 · similarity 0.89)
- Vision-Language Model Selection and Reuse for Downstream Adaptation (ICML 2025 · similarity 0.89)
- Efficient and Context-Aware Label Propagation for Zero-/Few-Shot Training-Free Adaptation of Vision-Language Model (ICLR 2025 · similarity 0.89)
- Machine Vision Therapy: Multimodal Large Language Models Can Enhance Visual Robustness via Denoising In-Context Learning (ICML 2024 · similarity 0.89)
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.
