ICLR 2023 accepted papers on Label-efficient and robust learning

Recognition and long-tail · 89 papers

• Surgical Fine-Tuning Improves Adaptation to Distribution Shifts (46 citations)
• Visual Recognition with Deep Nearest Centroids (45 citations)
• Post-hoc Concept Bottleneck Models (36 citations)
• Multitask Prompt Tuning Enables Parameter-Efficient Transfer Learning (30 citations)
• simpleKT: A Simple But Tough-to-Beat Baseline for Knowledge Tracing (26 citations)
• Optimizing Bi-Encoder for Named Entity Recognition via Contrastive Learning (22 citations)
• Code Translation with Compiler Representations (19 citations)
• Label-free Concept Bottleneck Models (17 citations)
• LPT: Long-tailed Prompt Tuning for Image Classification (17 citations)
• Unicom: Universal and Compact Representation Learning for Image Retrieval (17 citations)
• CUDA: Curriculum of Data Augmentation for Long-tailed Recognition (15 citations)
• TDR-CL: Targeted Doubly Robust Collaborative Learning for Debiased Recommendations (14 citations)

and 77 more in the interactive map.

Self-supervised and contrastive · 49 papers

• Protein Representation Learning by Geometric Structure Pretraining (40 citations)
• Masked Frequency Modeling for Self-Supervised Visual Pre-Training (29 citations)
• On the duality between contrastive and non-contrastive self-supervised learning (24 citations)
• Corrupted Image Modeling for Self-Supervised Visual Pre-Training (22 citations)
• Towards the Generalization of Contrastive Self-Supervised Learning (20 citations)
• Advancing Radiograph Representation Learning with Masked Record Modeling (18 citations)
• SimPer: Simple Self-Supervised Learning of Periodic Targets (18 citations)
• Transfer Learning with Deep Tabular Models (18 citations)
• BrainBERT: Self-supervised representation learning for intracranial recordings (14 citations)
• The hidden uniform cluster prior in self-supervised learning (13 citations)
• Hyperbolic Self-paced Learning for Self-supervised Skeleton-based Action Representations (11 citations)
• ImageNet-X: Understanding Model Mistakes with Factor of Variation Annotations (11 citations)

and 37 more in the interactive map.

Domain adaptation and generalization · 23 papers

• Divide to Adapt: Mitigating Confirmation Bias for Domain Adaptation of Black-Box Predictors (14 citations)
• Dirichlet-based Uncertainty Calibration for Active Domain Adaptation (11 citations)
• Sparse Mixture-of-Experts are Domain Generalizable Learners (11 citations)
• Information-Theoretic Analysis of Unsupervised Domain Adaptation (4 citations)
• Domain Generalisation via Domain Adaptation: An Adversarial Fourier Amplitude Approach (3 citations)
• ManyDG: Many-domain Generalization for Healthcare Applications (3 citations)
• Unsupervised Manifold Alignment with Joint Multidimensional Scaling (3 citations)
• Using Language to Extend to Unseen Domains (3 citations)
• Addressing Parameter Choice Issues in Unsupervised Domain Adaptation by Aggregation (2 citations)
• Domain-Indexing Variational Bayes: Interpretable Domain Index for Domain Adaptation (2 citations)
• Energy-Based Test Sample Adaptation for Domain Generalization (2 citations)
• Interpretations of Domain Adaptations via Layer Variational Analysis (2 citations)

and 11 more in the interactive map.

OOD and anomaly detection · 19 papers

• Pushing the Limits of Fewshot Anomaly Detection in Industry Vision: Graphcore (25 citations)
• Extremely Simple Activation Shaping for Out-of-Distribution Detection (23 citations)
• How to Exploit Hyperspherical Embeddings for Out-of-Distribution Detection? (23 citations)
• A Call to Reflect on Evaluation Practices for Failure Detection in Image Classification (16 citations)
• Broken Neural Scaling Laws (14 citations)
• Out-of-Distribution Detection and Selective Generation for Conditional Language Models (12 citations)
• Harnessing Out-Of-Distribution Examples via Augmenting Content and Style (10 citations)
• Out-of-distribution Detection with Implicit Outlier Transformation (10 citations)
• Pareto Invariant Risk Minimization: Towards Mitigating the Optimization Dilemma in Out-of-Distribution Generalization (10 citations)
• Agree to Disagree: Diversity through Disagreement for Better Transferability (7 citations)
• Non-parametric Outlier Synthesis (6 citations)
• A Closer Look at Model Adaptation using Feature Distortion and Simplicity Bias (4 citations)

and 7 more in the interactive map.

Zero-/few-shot and meta-learning · 17 papers

• Self-Guided Noise-Free Data Generation for Efficient Zero-Shot Learning (21 citations)
• Context-enriched molecule representations improve few-shot drug discovery (12 citations)
• Guess the Instruction! Flipped Learning Makes Language Models Stronger Zero-Shot Learners (10 citations)
• Prototypical Calibration for Few-shot Learning of Language Models (10 citations)
• Masked Unsupervised Self-training for Label-free Image Classification (7 citations)
• Model ensemble instead of prompt fusion: a sample-specific knowledge transfer method for few-shot prompt tuning (5 citations)
• Unsupervised Meta-learning via Few-shot Pseudo-supervised Contrastive Learning (5 citations)
• Cross-Level Distillation and Feature Denoising for Cross-Domain Few-Shot Classification (3 citations)
• Text Summarization with Oracle Expectation (3 citations)
• Universal Few-shot Learning of Dense Prediction Tasks with Visual Token Matching (3 citations)
• Contrastive Meta-Learning for Partially Observable Few-Shot Learning (2 citations)
• Massively Scaling Heteroscedastic Classifiers (1 citation)

and 5 more in the interactive map.

Semi-supervised learning · 15 papers

• FreeMatch: Self-adaptive Thresholding for Semi-supervised Learning (140 citations)
• Continuous pseudo-labeling from the start (6 citations)
• Imbalanced Semi-supervised Learning with Bias Adaptive Classifier (4 citations)
• SemPPL: Predicting Pseudo-Labels for Better Contrastive Representations (4 citations)
• STUNT: Few-shot Tabular Learning with Self-generated Tasks from Unlabeled Tables (3 citations)
• Confidence Estimation Using Unlabeled Data (2 citations)
• RoPAWS: Robust Semi-supervised Representation Learning from Uncurated Data (2 citations)
• Don’t fear the unlabelled: safe semi-supervised learning via debiasing (1 citation)
• InPL: Pseudo-labeling the Inliers First for Imbalanced Semi-supervised Learning
• Label Propagation with Weak Supervision
• Semi-supervised Community Detection via Structural Similarity Metrics
• Towards Better Selective Classification

and 3 more in the interactive map.

Noisy labels · 14 papers

• Free Lunch for Domain Adversarial Training: Environment Label Smoothing (24 citations)
• When Source-Free Domain Adaptation Meets Learning with Noisy Labels (21 citations)
• Mitigating Memorization of Noisy Labels via Regularization between Representations (8 citations)
• Noise-Robust De-Duplication at Scale (6 citations)
• Benign Overfitting in Classification: Provably Counter Label Noise with Larger Models (3 citations)
• Deep Learning From Crowdsourced Labels: Coupled Cross-Entropy Minimization, Identifiability, and Regularization (3 citations)
• Learning to Segment from Noisy Annotations: A Spatial Correction Approach (3 citations)
• Noise Injection Node Regularization for Robust Learning (2 citations)
• Over-Training with Mixup May Hurt Generalization (2 citations)
• The Curious Case of Benign Memorization (1 citation)
• A law of adversarial risk, interpolation, and label noise
• Leveraging Unlabeled Data to Track Memorization

and 2 more in the interactive map.

Knowledge distillation · 13 papers

• Better Teacher Better Student: Dynamic Prior Knowledge for Knowledge Distillation (19 citations)
• HomoDistil: Homotopic Task-Agnostic Distillation of Pre-trained Transformers (9 citations)
• IDEAL: Query-Efficient Data-Free Learning from Black-Box Models (5 citations)
• Self-Distillation for Further Pre-training of Transformers (4 citations)
• Function-Consistent Feature Distillation (3 citations)
• Augmentation with Projection: Towards an Effective and Efficient Data Augmentation Paradigm for Distillation (2 citations)
• Meta-prediction Model for Distillation-Aware NAS on Unseen Datasets (2 citations)
• Robust Active Distillation (2 citations)
• Revisiting the Entropy Semiring for Neural Speech Recognition (1 citation)
• Supervision Complexity and its Role in Knowledge Distillation (1 citation)
• Teacher Guided Training: An Efficient Framework for Knowledge Transfer (1 citation)
• Scaffolding a Student to Instill Knowledge

and 1 more in the interactive map.

Test-time adaptation · 9 papers

• Towards Stable Test-time Adaptation in Dynamic Wild World (64 citations)
• TTN: A Domain-Shift Aware Batch Normalization in Test-Time Adaptation (21 citations)
• Test-Time Adaptation via Self-Training with Nearest Neighbor Information (12 citations)
• DELTA: DEGRADATION-FREE FULLY TEST-TIME ADAPTATION (8 citations)
• A Learning Based Hypothesis Test for Harmful Covariate Shift (3 citations)
• First Steps Toward Understanding the Extrapolation of Nonlinear Models to Unseen Domains (3 citations)
• Towards Understanding GD with Hard and Conjugate Pseudo-labels for Test-Time Adaptation (2 citations)
• Distributionally Robust Post-hoc Classifiers under Prior Shifts
• M-L2O: Towards Generalizable Learning-to-Optimize by Test-Time Fast Self-Adaptation