Research map / Label-efficient and robust learning / Noisy labels
Noise Is Also Useful: Negative Correlation-Steered Latent Contrastive Learning
Jiexi Yan, Lei Luo, Chenghao Xu, Cheng Deng, Heng Huang
This paper tried a common change: Add a contrastive loss
Add latent contrastive loss · CIFAR-100, symmetric noise 40% · test accuracy (%): 59.12 -> 68.59 Improved
What 1,317 accepted papers report for this change
- 1,253 improved
- 22 no clear effect
- 16 mixed
- 26 hurt
- Beyond Looking Up, Try Looking Around: Harmonizing Global Structure and Local Consistency in Optimal Transport for Short Text Clustering ICML 2026 · add contrastive learning · SearchSnippets · ACC: 56.65 -> 80.12Improved
- SleepLM: Natural-Language Intelligence for Human Sleep ICML 2026 · add contrastive objective to captioning (CoCa vs Cap) · sleep event IoU · IoU: 29.2 -> 27.4Hurt
- LIMSSR: LLM-Driven Sequence-to-Score Reasoning under Training-Time Incomplete Multimodal Observations ICML 2026 · add contrastive loss L_con · FS1000 average · Spearman correlation: 0.759 -> 0.789Improved
See this change in detailMatched by the wording of each paper's own result, read by AI from the full text. Papers report gains more often than failures, so this shows what gets published, not the chance a change works.
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 LaCoL instead of JNPL | CIFAR-10 with symmetric label noise · CIFAR-10, symmetric noise 20%–60% | test accuracy (%): 93.53 -> 94.12; 91.89 -> 92.33; 88.45 -> 88.72 | Improved |
| Use LaCoL instead of JNPL | CIFAR-10 with asymmetric label noise · CIFAR-10, asymmetric noise 20%–40% | test accuracy (%): 93.45 -> 93.76; 90.72 -> 91.07 | Improved |
| Use LaCoL instead of the strongest listed baseline | CIFAR-100 with symmetric label noise · CIFAR-100, symmetric noise 20%–60% | test accuracy (%): 71.75 -> 71.24; 68.11 -> 68.59; 61.26 -> 61.93 | Mixed |
| Use LaCoL instead of the strongest listed baseline | CIFAR-100 with asymmetric label noise · CIFAR-100, asymmetric noise 20%–40% | test accuracy (%): 70.13 -> 70.84; 59.51 -> 59.73 | Improved |
| Use LaCoL instead of the strongest listed baseline | CIFAR-10 with instance-dependent noise · CIFAR-10, instance-dependent noise 20%–40% | test accuracy (%): 90.41 -> 92.47; 84.78 -> 86.76 | Improved |
| Use LaCoL instead of CDR | CIFAR-100 with instance-dependent noise · CIFAR-100, instance-dependent noise 20%–40% | test accuracy (%): 67.33 -> 70.16; 55.94 -> 63.87 | Improved |
| Use LaCoL instead of JNPL | Clothing-1M · Clothing-1M | test accuracy (%): 74.15 -> 74.68 | Improved |
| Add diversity-preserving weak-selection/strong-training augmentation (ablation) | LaCoL on CIFAR-100 · CIFAR-100, symmetric noise 40% | test accuracy (%): 67.24 -> 68.59 | Improved |
| Add diversity-preserving weak-selection/strong-training augmentation (ablation) | LaCoL on CIFAR-100 · CIFAR-100, asymmetric noise 40% | test accuracy (%): 58.22 -> 59.73 | Improved |
| Add latent contrastive loss (ablation) | LaCoL on CIFAR-100 · CIFAR-100, symmetric noise 40% | test accuracy (%): 59.12 -> 68.59 | Improved |
| Add latent contrastive loss (ablation) | LaCoL on CIFAR-100 · CIFAR-100, asymmetric noise 40% | test accuracy (%): 47.71 -> 59.73 | Improved |
| Add cross-space similarity consistency loss (ablation) | LaCoL on CIFAR-100 · CIFAR-100, symmetric noise 40% | test accuracy (%): 66.91 -> 68.59 | Improved |
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.
- Co-teaching+ — Co-teaching: Robust training of deep neural networks with extremely noisy labels (NeurIPS 2018) · 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
- FINE Samples for Learning with Noisy Labels (NeurIPS 2021 · similarity 0.89)
- On Learning Contrastive Representations for Learning With Noisy Labels (CVPR 2022 · similarity 0.89)
- Multi-Objective Interpolation Training for Robustness To Label Noise (CVPR 2021 · similarity 0.88)
- Fine-Grained Classification With Noisy Labels (CVPR 2023 · similarity 0.88)
- Adaptive Integration of Partial Label Learning and Negative Learning for Enhanced Noisy Label Learning (AAAI 2024 · similarity 0.87)
- Learning with Structural Labels for Learning with Noisy Labels (CVPR 2024 · 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.
