ICLR 2025 accepted papers on Label-efficient and robust learning

Recognition and long-tail · 93 papers

• A Decade's Battle on Dataset Bias: Are We There Yet? (8 citations)
• A new framework for evaluating model out-of-distribution generalisation for the biochemical domain (5 citations)
• BirdSet: A Large-Scale Dataset for Audio Classification in Avian Bioacoustics (4 citations)
• Contextual Document Embeddings (3 citations)
• Immunogenicity Prediction with Dual Attention Enables Vaccine Target Selection (3 citations)
• Learning to engineer protein flexibility (2 citations)
• ContextGNN: Beyond Two-Tower Recommendation Systems (1 citation)
• Differentiable Optimization of Similarity Scores Between Models and Brains (1 citation)
• Enhancing End-to-End Autonomous Driving with Latent World Model (1 citation)
• Feature Responsiveness Scores: Model-Agnostic Explanations for Recourse (1 citation)
• I-Con: A Unifying Framework for Representation Learning (1 citation)
• Intrinsic User-Centric Interpretability through Global Mixture of Experts (1 citation)

and 81 more in the interactive map.

Self-supervised and contrastive · 35 papers

• Distilling Structural Representations into Protein Sequence Models (11 citations)
• PaPaGei: Open Foundation Models for Optical Physiological Signals (5 citations)
• Cross-Entropy Is All You Need To Invert the Data Generating Process (1 citation)
• T-JEPA: Augmentation-Free Self-Supervised Learning for Tabular Data (1 citation)
• Universal Image Restoration Pre-training via Degradation Classification (1 citation)
• UNSURE: self-supervised learning with Unknown Noise level and Stein's Unbiased Risk Estimate (1 citation)
• $\mathbb{X}$-Sample Contrastive Loss: Improving Contrastive Learning with Sample Similarity Graphs
• A Theoretical Analysis of Self-Supervised Learning for Vision Transformers
• A Theoretical Perspective: How to Prevent Model Collapse in Self-consuming Training Loops
• Beyond Random Augmentations: Pretraining with Hard Views
• Can a MISL Fly? Analysis and Ingredients for Mutual Information Skill Learning
• CircuitFusion: Multimodal Circuit Representation Learning for Agile Chip Design

and 23 more in the interactive map.

OOD and anomaly detection · 27 papers

• In Search of Forgotten Domain Generalization (2 citations)
• Adversarially Robust Out-of-Distribution Detection Using Lyapunov-Stabilized Embeddings (1 citation)
• Outlier Synthesis via Hamiltonian Monte Carlo for Out-of-Distribution Detection (1 citation)
• The Pitfalls of Memorization: When Memorization Hurts Generalization (1 citation)
• Advancing Out-of-Distribution Detection via Local Neuroplasticity
• Benign Overfitting in Out-of-Distribution Generalization of Linear Models
• Can In-context Learning Really Generalize to Out-of-distribution Tasks?
• Forte : Finding Outliers with Representation Typicality Estimation
• Intermediate Layer Classifiers for OOD generalization
• Local-Prompt: Extensible Local Prompts for Few-Shot Out-of-Distribution Detection
• MetaOOD: Automatic Selection of OOD Detection Models
• One-for-All Few-Shot Anomaly Detection via Instance-Induced Prompt Learning

and 15 more in the interactive map.

Knowledge distillation · 26 papers

• Speculative Knowledge Distillation: Bridging the Teacher-Student Gap Through Interleaved Sampling (2 citations)
• MiniPLM: Knowledge Distillation for Pre-training Language Models (1 citation)
• Accessing Vision Foundation Models via ImageNet-1K
• BLEND: Behavior-guided Neural Population Dynamics Modeling via Privileged Knowledge Distillation
• Breaking Class Barriers: Efficient Dataset Distillation via Inter-Class Feature Compensator
• Dataset Distillation via Knowledge Distillation: Towards Efficient Self-Supervised Pre-training of Deep Networks
• Elucidating the Preconditioning in Consistency Distillation
• Gap Preserving Distillation by Building Bidirectional Mappings with A Dynamic Teacher
• Going Beyond Feature Similarity: Effective Dataset distillation based on Class-aware Conditional Mutual Information
• High-dimensional Analysis of Knowledge Distillation: Weak-to-Strong Generalization and Scaling Laws
• Knowledge Distillation with Multi-granularity Mixture of Priors for Image Super-Resolution
• Montessori-Instruct: Generate Influential Training Data Tailored for Student Learning

and 14 more in the interactive map.

Test-time adaptation · 18 papers

• Adaptive Camera Sensor for Vision Models
• Addressing Label Shift in Distributed Learning via Entropy Regularization
• COME: Test-time Adaption by Conservatively Minimizing Entropy
• Learning Neural Networks with Distribution Shift: Efficiently Certifiable Guarantees
• Matcha: Mitigating Graph Structure Shifts with Test-Time Adaptation
• Multi-Label Test-Time Adaptation with Bound Entropy Minimization
• Noisy Test-Time Adaptation in Vision-Language Models
• On the Adversarial Risk of Test Time Adaptation: An Investigation into Realistic Test-Time Data Poisoning
• Smoothing the Shift: Towards Stable Test-Time Adaptation under Complex Multimodal Noises
• Test-Time Adaptation for Combating Missing Modalities in Egocentric Videos
• Test-time Adaptation for Image Compression with Distribution Regularization
• Test-time Adaptation for Regression by Subspace Alignment

and 6 more in the interactive map.

Domain adaptation and generalization · 12 papers

• Proxy Denoising for Source-Free Domain Adaptation (1 citation)
• Denoising as Adaptation: Noise-Space Domain Adaptation for Image Restoration
• DynAlign: Unsupervised Dynamic Taxonomy Alignment for Cross-Domain Segmentation
• HiLo: A Learning Framework for Generalized Category Discovery Robust to Domain Shifts
• Is Large-scale Pretraining the Secret to Good Domain Generalization?
• Multimodal Unsupervised Domain Generalization by Retrieving Across the Modality Gap
• SPDIM: Source-Free Unsupervised Conditional and Label Shift Adaptation in EEG
• Towards Domain Adaptive Neural Contextual Bandits
• Cross-Domain Offline Policy Adaptation with Optimal Transport and Dataset Constraint
• Decoupled Finetuning for Domain Generalizable Semantic Segmentation
• Revisiting Source-Free Domain Adaptation: a New Perspective via Uncertainty Control
• Selective Unlearning via Representation Erasure Using Domain Adversarial Training

Zero-/few-shot and meta-learning · 11 papers

• ZeroDiff: Solidified Visual-semantic Correlation in Zero-Shot Learning (2 citations)
• Learning Clustering-based Prototypes for Compositional Zero-Shot Learning (1 citation)
• LR0.FM: Low-Res Benchmark and Improving robustness for Zero-Shot Classification in Foundation Models
• Manifold Induced Biases for Zero-shot and Few-shot Detection of Generated Images
• PharmacoMatch: Efficient 3D Pharmacophore Screening via Neural Subgraph Matching
• Rapidly Adapting Policies to the Real-World via Simulation-Guided Fine-Tuning
• Sensor-Invariant Tactile Representation
• SynQ: Accurate Zero-shot Quantization by Synthesis-aware Fine-tuning
• UniMatch: Universal Matching from Atom to Task for Few-Shot Drug Discovery
• Unsupervised Meta-Learning via In-Context Learning
• Why In-Context Learning Models are Good Few-Shot Learners?

Semi-supervised learning · 5 papers

• Re-Evaluating the Impact of Unseen-Class Unlabeled Data on Semi-Supervised Learning Model
• Towards Understanding Why FixMatch Generalizes Better Than Supervised Learning
• Gaussian-Based Instance-Adaptive Intensity Modeling for Point-Supervised Facial Expression Spotting
• Supervised and Semi-Supervised Diffusion Maps with Label-Driven Diffusion
• Towards Unbiased Learning in Semi-Supervised Semantic Segmentation

Noisy labels · 3 papers

• GIFT: Unlocking Full Potential of Labels in Distilled Dataset at Near-zero Cost
• Regretful Decisions under Label Noise
• Noise Separation guided Candidate Label Reconstruction for Noisy Partial Label Learning