Research map / Label-efficient and robust learning / Noisy labels
Mask-guided Semantic Alignment: Robust Learning with Noisy Labels via Temporal Attention Stability
Yubo Nian, Can Gao
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 ATC (attention temporal consistency) sample selection (ablation) | baseline LNL framework · CIFAR-10 (sym 90% noise) | test accuracy: 86.8 -> 94.6 | Improved |
| add ATC sample selection (ablation) | baseline LNL framework · CIFAR-100 (sym 90% noise) | test accuracy: 44.0 -> 49.6 | Improved |
| add ATC sample selection (ablation) | baseline LNL framework · CIFAR-10 (sym 20% noise) | test accuracy: 97.1 -> 97.1 | No clear effect |
| add AAM (adaptive attention masks) on top of ATC (ablation) | ATC model · CIFAR-100 (sym 80% noise) | test accuracy: 70.5 -> 72.3 | Improved |
| add AAM on top of ATC (ablation) | ATC model · CIFAR-100 (sym 90% noise) | test accuracy: 49.6 -> 53.3 | Improved |
| add MFA (mask-guided feature alignment) on top of ATC+AAM (ablation) | ATC+AAM model · CIFAR-100 (sym 90% noise) | test accuracy: 53.3 -> 55.6 | Improved |
| add MFA on top of ATC+AAM (ablation) | ATC+AAM model · CIFAR-100 (sym 80% noise) | test accuracy: 72.3 -> 73.3 | Improved |
| add MFA on top of ATC+AAM (ablation) | ATC+AAM model · CIFAR-10 (sym 90% noise) | test accuracy: 94.8 -> 95.1 | Improved |
| use full method instead of DULC under extreme noise | Ours vs DULC · CIFAR-100 (sym 90% noise) | test accuracy: 52.8 -> 55.6 | Improved |
| use full method instead of DULC under extreme noise | Ours vs DULC · CIFAR-10 (sym 90% noise) | test accuracy: 93.5 -> 95.1 | Improved |
| use full method instead of best baseline CCLRL | Ours vs CCLRL · Animal-10N | accuracy: 89.7 -> 90.8 | Improved |
| vary mask ratio mu (optimum at 0.6) (ablation) | AAM · CIFAR-100 (sym 80% noise) | test accuracy: peaks at 73.3% when mu = 0.6 | Mixed |
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.
- SURE — SURE: SUrvey REcipes for building reliable and robust deep networks (CVPR 2024) · better on 1 of 1 shared result
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
- Mitigating Endogenous Confirmation Bias in Noisy Label Learning for Vision-Language Models (AAAI 2026 · similarity 0.88)
- SyncMask: Synchronized Attentional Masking for Fashion-centric Vision-Language Pretraining (CVPR 2024 · similarity 0.87)
- DH-Set: Improving Vision-Language Alignment with Diverse and Hybrid Set-Embeddings Learning (CVPR 2025 · similarity 0.86)
- Fine-Grained Verifiers: Preference Modeling as Next-token Prediction in Vision-Language Alignment (ICLR 2025 · similarity 0.86)
- Pre-Trained Vision-Language Models as Noisy Partial Annotators (AAAI 2025 · similarity 0.86)
- Enhance Vision-Language Alignment with Noise (AAAI 2025 · similarity 0.86)
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.
