ICLR 2024 accepted papers on Label-efficient and robust learning
221 of the 2,259 papers accepted at ICLR 2024 (9.8%) are on Label-efficient and robust learning, compared with 15.7% at ICLR 2023. They fall into 9 topics.
See ICLR 2024 in the interactive map
Working on something in this area? Describe your idea in scime atlas to see the closest accepted papers by meaning and how crowded the spot has become.
Topics and papers
Recognition and long-tail · 89 papers
- From Molecules to Materials: Pre-training Large Generalizable Models for Atomic Property Prediction (22 citations)
- Geographic Location Encoding with Spherical Harmonics and Sinusoidal Representation Networks (11 citations)
- TabR: Tabular Deep Learning Meets Nearest Neighbors (9 citations)
- Yet Another ICU Benchmark: A Flexible Multi-Center Framework for Clinical ML (8 citations)
- Task structure and nonlinearity jointly determine learned representational geometry (7 citations)
- Learning to design protein-protein interactions with enhanced generalization (6 citations)
- On the Foundations of Shortcut Learning (6 citations)
- Real-Fake: Effective Training Data Synthesis Through Distribution Matching (6 citations)
- Towards Foundational Models for Molecular Learning on Large-Scale Multi-Task Datasets (6 citations)
- AdaMerging: Adaptive Model Merging for Multi-Task Learning (5 citations)
- Energy-Based Concept Bottleneck Models: Unifying Prediction, Concept Intervention, and Probabilistic Interpretations (4 citations)
- How Well Do Supervised 3D Models Transfer to Medical Imaging Tasks? (4 citations)
and 77 more in the interactive map.
Self-supervised and contrastive · 37 papers
- MERT: Acoustic Music Understanding Model with Large-Scale Self-supervised Training (28 citations)
- Large-scale Training of Foundation Models for Wearable Biosignals (23 citations)
- Exploring Target Representations for Masked Autoencoders (16 citations)
- MOTOR: A Time-to-Event Foundation Model For Structured Medical Records (10 citations)
- HypeBoy: Generative Self-Supervised Representation Learning on Hypergraphs (4 citations)
- Protein-ligand binding representation learning from fine-grained interactions (4 citations)
- Investigating the Benefits of Projection Head for Representation Learning (3 citations)
- Memorization in Self-Supervised Learning Improves Downstream Generalization (3 citations)
- Self-supervised Pocket Pretraining via Protein Fragment-Surroundings Alignment (3 citations)
- Self-supervised Representation Learning from Random Data Projectors (3 citations)
- CODE REPRESENTATION LEARNING AT SCALE (1 citation)
- LiDAR: Sensing Linear Probing Performance in Joint Embedding SSL Architectures (1 citation)
and 25 more in the interactive map.
Domain adaptation and generalization · 24 papers
- BayesPrompt: Prompting Large-Scale Pre-Trained Language Models on Few-shot Inference via Debiased Domain Abstraction (3 citations)
- ReSimAD: Zero-Shot 3D Domain Transfer for Autonomous Driving with Source Reconstruction and Target Simulation (3 citations)
- SF(DA)$^2$: Source-free Domain Adaptation Through the Lens of Data Augmentation (3 citations)
- Adapting to Distribution Shift by Visual Domain Prompt Generation (2 citations)
- Alice Benchmarks: Connecting Real World Re-Identification with the Synthetic (2 citations)
- Improving Domain Generalization with Domain Relations (2 citations)
- AUGCAL: Improving Sim2Real Adaptation by Uncertainty Calibration on Augmented Synthetic Images (1 citation)
- Can We Evaluate Domain Adaptation Models Without Target-Domain Labels? (1 citation)
- Continuous Invariance Learning (1 citation)
- Source-Free and Image-Only Unsupervised Domain Adaptation for Category Level Object Pose Estimation (1 citation)
- Tight Rates in Supervised Outlier Transfer Learning (1 citation)
- Towards Identifiable Unsupervised Domain Translation: A Diversified Distribution Matching Approach (1 citation)
and 12 more in the interactive map.
OOD and anomaly detection · 21 papers
- Neuron Activation Coverage: Rethinking Out-of-distribution Detection and Generalization (4 citations)
- Does CLIP’s generalization performance mainly stem from high train-test similarity? (3 citations)
- NECO: NEural Collapse Based Out-of-distribution detection (3 citations)
- DOS: Diverse Outlier Sampling for Out-of-Distribution Detection (2 citations)
- Out-of-Distribution Detection by Leveraging Between-Layer Transformation Smoothness (2 citations)
- Scaling for Training Time and Post-hoc Out-of-distribution Detection Enhancement (2 citations)
- Spurious Feature Diversification Improves Out-of-distribution Generalization (2 citations)
- How Does Unlabeled Data Provably Help Out-of-Distribution Detection? (1 citation)
- HYPO: Hyperspherical Out-Of-Distribution Generalization (1 citation)
- ImageNet-OOD: Deciphering Modern Out-of-Distribution Detection Algorithms (1 citation)
- Learning with Mixture of Prototypes for Out-of-Distribution Detection (1 citation)
- Overcoming the Pitfalls of Vision-Language Model Finetuning for OOD Generalization (1 citation)
and 9 more in the interactive map.
Zero-/few-shot and meta-learning · 16 papers
- MuSc: Zero-Shot Industrial Anomaly Classification and Segmentation with Mutual Scoring of the Unlabeled Images (7 citations)
- Context-Aware Meta-Learning (3 citations)
- Quick-Tune: Quickly Learning Which Pretrained Model to Finetune and How (3 citations)
- SPTNet: An Efficient Alternative Framework for Generalized Category Discovery with Spatial Prompt Tuning (3 citations)
- BECLR: Batch Enhanced Contrastive Few-Shot Learning (2 citations)
- Understanding the Robustness of Multi-modal Contrastive Learning to Distribution Shift (2 citations)
- A Hard-to-Beat Baseline for Training-free CLIP-based Adaptation (1 citation)
- Gen-Z: Generative Zero-Shot Text Classification with Contextualized Label Descriptions (1 citation)
- MetaCoCo: A New Few-Shot Classification Benchmark with Spurious Correlation (1 citation)
- Towards Few-Shot Adaptation of Foundation Models via Multitask Finetuning (1 citation)
- Zero-Shot Robustification of Zero-Shot Models (1 citation)
- Constrained Decoding for Cross-lingual Label Projection
and 4 more in the interactive map.
Knowledge distillation · 13 papers
- Knowledge Distillation Based on Transformed Teacher Matching (5 citations)
- Towards Lossless Dataset Distillation via Difficulty-Aligned Trajectory Matching (4 citations)
- On-Policy Distillation of Language Models: Learning from Self-Generated Mistakes (3 citations)
- Hybrid Distillation: Connecting Masked Autoencoders with Contrastive Learners (2 citations)
- Less or More From Teacher: Exploiting Trilateral Geometry For Knowledge Distillation (2 citations)
- AuG-KD: Anchor-Based Mixup Generation for Out-of-Domain Knowledge Distillation (1 citation)
- Bayes Conditional Distribution Estimation for Knowledge Distillation Based on Conditional Mutual Information (1 citation)
- Toward Student-oriented Teacher Network Training for Knowledge Distillation (1 citation)
- Ensemble Distillation for Unsupervised Constituency Parsing
- Fantastic Gains and Where to Find Them: On the Existence and Prospect of General Knowledge Transfer between Any Pretrained Model
- Towards the Fundamental Limits of Knowledge Transfer over Finite Domains
- A Good Learner can Teach Better: Teacher-Student Collaborative Knowledge Distillation
and 1 more in the interactive map.
Test-time adaptation · 8 papers
- On the Variance of Neural Network Training with respect to Test Sets and Distributions (7 citations)
- Entropy is not Enough for Test-Time Adaptation: From the Perspective of Disentangled Factors (5 citations)
- Fusing Models with Complementary Expertise (2 citations)
- Active Test-Time Adaptation: Theoretical Analyses and An Algorithm (1 citation)
- Un-Mixing Test-Time Normalization Statistics: Combatting Label Temporal Correlation
- Learning model uncertainty as variance-minimizing instance weights
- PROGRAM: PROtotype GRAph Model based Pseudo-Label Learning for Test-Time Adaptation
- Test-time Adaptation against Multi-modal Reliability Bias
Noisy labels · 7 papers
- Understanding and Mitigating the Label Noise in Pre-training on Downstream Tasks (5 citations)
- Early Stopping Against Label Noise Without Validation Data (1 citation)
- Unraveling the Enigma of Double Descent: An In-depth Analysis through the Lens of Learned Feature Space (1 citation)
- Dirichlet-based Per-Sample Weighting by Transition Matrix for Noisy Label Learning
- Understanding Domain Generalization: A Noise Robustness Perspective
- Why is SAM Robust to Label Noise?
- Robust Classification via Regression for Learning with Noisy Labels
Semi-supervised learning · 6 papers
- Robust Training of Federated Models with Extremely Label Deficiency (5 citations)
- SemiReward: A General Reward Model for Semi-supervised Learning (4 citations)
- Multimodal Learning Without Labeled Multimodal Data: Guarantees and Applications (1 citation)
- Out-Of-Domain Unlabeled Data Improves Generalization
- Realistic Evaluation of Semi-supervised Learning Algorithms in Open Environments
- Variance-enlarged Poisson Learning for Graph-based Semi-Supervised Learning with Extremely Sparse Labeled Data