Learning-with-Noisy-Labels

A curated list of most recent papers & codes in Learning with Noisy Labels

Some recent works about group-distributional robustness, label distribution shifts, are also included.

Public Software

Docta-AI: An advanced data-centric AI platform that detects and rectifies issues in any data format (i.e., label error detection). [Website]

Competition

A Hands-on Tutorial for Learning with Noisy Labels (IJCAI 2022)[website]

Tutorial

1st Learning and Mining with Noisy Labels Challenge (IJCAI 2023)[Website][GitHub]

Content

• Benchmarks & Leaderboard
• Papers & Code in 2023
  • KDD 2023
  • NeurIPS 2023
  • ICML 2023
  • CVPR 2023
  • ICLR 2023
• Papers & Code in 2022
  • NeurIPS 2022
  • ECCV 2022
  • ICML 2022
  • CVPR 2022
  • ICLR 2022
  • AISTATS 2022
  • Other Conferences 2022
  • ArXiv 2022
• Papers & Code in 2021
  • NeurIPS 2021
  • IJCAI 2021
  • ICML 2021
  • ICLR 2021
  • CVPR 2021
  • AISTATS 2021
  • AAAI 2021
  • Other Conferences 2021
  • ArXiv 2021
• Papers & Code in 2020
  • ICML 2020
  • ICLR 2020
  • Nips 2020
  • AAAI 2020
  • CVPR 2020
  • ECCV 2020
  • ArXiv 2020

Benchmarks & Leaderboard

Real-world noisy-label bechmarks:

Simulation of label noise: An Instance-Dependent Simulation Framework for Learning with Label Noise. [Paper]

This repo focus on papers after 2019, for previous works, please refer to (https://github.com/subeeshvasu/Awesome-Learning-with-Label-Noise).

Papers & Code in 2023

KDD 2023

• [UCSC REAL Lab] To Aggregate or Not? Learning with Separate Noisy Labels. [Paper]
• DyGen: Learning from Noisy Labels via Dynamics-Enhanced Generative Modeling. [Paper][Code]
• Robust Positive-Unlabeled Learning via Noise Negative Sample Self-correction. [Paper]
• Neural-Hidden-CRF: A Robust Weakly-Supervised Sequence Labeler. [Paper][Code]
• Complementary Classifier Induced Partial Label Learning. [Paper][Code]
• Partial-label Learning with Mixed Closed-Set and Open-Set Out-of-Candidate Examples. [Paper]
• Weakly Supervised Multi-Label Classification of Full-Text Scientific Papers. [Paper][Code]

NeurIPS 2023

• The Pursuit of Human Labeling: A New Perspective on Unsupervised Learning. [Paper][Code]
• AQuA: A Benchmarking Tool for Label Quality Assessment. [Paper]
• Efficient Testable Learning of Halfspaces with Adversarial Label Noise. [Paper]
• Neural Relation Graph: A Unified Framework for Identifying Label Noise and Outlier Data. [Paper][Code]
• Robust Data Pruning under Label Noise via Maximizing Re-labeling Accuracy. [Paper]
• Subclass-Dominant Label Noise: A Counterexample for the Success of Early Stopping. [Paper][Code]
• Label Correction of Crowdsourced Noisy Annotations with an Instance-Dependent Noise Transition Model. [Paper]
• Scale-teaching: Robust Multi-scale Training for Time Series Classification with Noisy Labels. [Paper][Code]
• SoTTA: Robust Test-Time Adaptation on Noisy Data Streams. [Paper][Code]
• Active Negative Loss Functions for Learning with Noisy Labels. [Paper][Code]
• Label-Retrieval-Augmented Diffusion Models for Learning from Noisy Labels. [Paper][Code]
• Training shallow ReLU networks on noisy data using hinge loss: when do we overfit and is it benign? [Paper]
• CSOT: Curriculum and Structure-Aware Optimal Transport for Learning with Noisy Labels. [Paper][Code]
• Deep Insights into Noisy Pseudo Labeling on Graph Data. [Paper]
• ARTIC3D: Learning Robust Articulated 3D Shapes from Noisy Web Image Collections. [Paper][Code]
• ALIM: Adjusting Label Importance Mechanism for Noisy Partial Label Learning. [Paper][Code]
• Weakly-Supervised Concealed Object Segmentation with SAM-based Pseudo Labeling and Multi-scale Feature Grouping. [Paper][Code]
• Label Poisoning is All You Need. [Paper][Code]
• SLaM: Student-Label Mixing for Distillation with Unlabeled Examples. [Paper]
• IPMix: Label-Preserving Data Augmentation Method for Training Robust Classifiers. [Paper]
• HQA-Attack: Toward High Quality Black-Box Hard-Label Adversarial Attack on Text. [Paper][Code]

ICML 2023

• [UCSC REAL Lab] Identifiability of Label Noise Transition Matrix. [Paper]
• Which is Better for Learning with Noisy Labels: The Semi-supervised Method or Modeling Label Noise? [Paper]
• Mitigating Memorization of Noisy Labels by Clipping the Model Prediction. [Paper][Code]
• CrossSplit: Mitigating Label Noise Memorization through Data Splitting. [Paper][Code]
• Understanding Self-Distillation in the Presence of Label Noise. [Paper]
• RandomClassificationNoisedoesnotdefeatAllConvexPotentialBoosters IrrespectiveofModelChoice. [Paper]
• Deep Clustering with Incomplete Noisy Pairwise Annotations: A Geometric Regularization Approach. [Paper]
• Delving into Noisy Label Detection with Clean Data. [Paper]
• When does Privileged information Explain Away Label Noise? [Paper]
• Squeeze, Recover and Relabel: Dataset Condensation at ImageNet Scale From A New Perspective. [Paper][Code]
• Promises and Pitfalls of Threshold-based Auto-labeling. [Paper]
• Accelerating Exploration with Unlabeled Prior Data. [Paper]

CVPR 2023

• Twin Contrastive Learning with Noisy Labels. [Paper][Code]
• Exploring High-Quality Pseudo Masks for Weakly Supervised Instance Segmentation. [Paper][Code]
• HandsOff: Labeled Dataset Generation with No Additional Human Annotations. [Paper][Code]
• Learning from Noisy Labels with Decoupled Meta Label Purifier. [Paper][Code]
• DISC: Learning from Noisy Labels via Dynamic Instance-Specific Selection and Correction. [Paper][Code]
• Leveraging Inter-Rater Agreement for Classification in the Presence of Noisy Labels. [Paper]
• Fine-Grained Classification with Noisy Labels. [Paper]
• Collaborative Noisy Label Cleaner: Learning Scene-aware Trailers for Multi-modal Highlight Detection in Movies. [Paper][Code]
• MixTeacher: Mining Promising Labels with Mixed Scale Teacher for Semi-supervised Object Detection. [Paper][Code]
• OT-Filter: An Optimal Transport Filter for Learning With Noisy Labels. [Paper]
• Exploiting Completeness and Uncertainty of Pseudo Labels for Weakly Supervised Video Anomaly Detection. [Paper][Code]
• Semi-Supervised 2D Human Pose Estimation Driven by Position Inconsistency Pseudo Label Correction Module. [Paper][Code]
• Learning with Noisy labels via Self-supervised