Study shows scaling up models doesn't always improve downstream tasks.
problem Understanding why scaling up models doesn't always improve downstream performance.
method Systematic study of 4800 experiments on various models, analyzing performance on 20 downstream tasks.
result Performance on downstream tasks saturates as model size increases, revealing a nonlinear relationship.
New insights into contrastive learning reveal how projectors affect downstream performance.
problem Understanding how projectors in contrastive learning impact downstream linear classification accuracy.
method Identified and modeled two effects: expansion and shrinkage induced by contrastive loss.
result Linear projectors operating in the shrinkage regime hinder downstream classification accuracy.
New algorithm balances spatial data approximation and prediction accuracy.
problem Lack of methods considering spatial correlation and downstream modeling in dimension reduction.
method Formalizes approximation and modeling utility as metrics, proposes a balanced algorithm.
result Optimal trade-off between approximation accuracy and downstream modeling utility.
New measure assesses time series pre-training data quality without labels.
problem Challenges in collecting diverse pre-training datasets for time series classification.
method Contrastive-learning-based foundation model and contrastive accuracy measure.
result Contrastive accuracy correlates with model performance on downstream tasks.
This study examines how the size and alignment of pretraining data affect the performance of large language models on downstream tasks.
problem Understanding how the size and alignment of pretraining data impact the performance of large language models on downstream tasks.
method Investigated the scaling behavior of large language models in a transfer learning setting, focusing on machine translation tasks.
result The size of the finetuning dataset and the distribution alignment between pretraining and downstream data significantly influence the scaling behavior of downstream performance.
Self-supervised metric learning boosts downstream tasks in multi-view data.
problem Improving distance-based downstream tasks without labeled data.
method Developed a statistical framework to study self-supervised metric learning in multi-view data.
result Self-supervised metric learning improves target distances for various downstream tasks.
Study improves healthcare time series imputation by considering structured missingness.
problem Structured missingness in clinical data impacts time series imputation models.
method Analysis of different masking strategies on imputation methods using PhysioNet Challenge 2012 dataset.
result Masking choices significantly affect imputation accuracy and clinical prediction.
This work investigates how neural collapse improves transfer learning for large-scale models.
problem Improving transfer learning for large-scale models with limited labeled data.
method Investigates neural collapse and develops a fine-tuning method using skip-connections.
result Feature collapse on downstream data correlates with higher transfer accuracy.
Adversarially robust models transfer better than standard models in image classification.
problem Improving transfer learning performance in image classification.
method Focused on adversarially robust ImageNet classifiers, compared to standard models.
result Adversarially robust models yield improved accuracy on downstream classification tasks.
Improved tabular models learn better from real-world data.
problem Tabular models perform poorly on real-world datasets when trained only on synthetic data.
method Continued pre-training on a curated set of real-world datasets.
result Real-TabPFN achieves superior predictive accuracy on 29 datasets.
CROP verifies clean prefixes in reasoning traces, improving downstream repair accuracy.
problem Uncertainty in reasoning traces prevents full certification of entire responses.
method CROP selects a calibrated threshold to certify the longest prefix with low risk proxies.
result CROP improves downstream repair accuracy by preserving valid reasoning and discarding misleading suffixes.
Prototype model improves model auditing and understanding.
problem Auditing and understanding modern language models is expensive and approximate.
method Introduced a sparse, non-negative mixture of learned prototypes trained with clustering objectives.
result Prototype models either surpass or remain within 2.5 percentage points of dense baselines on downstream tasks.
This work defines idealized SSL representations and improves existing methods.
problem Unclear characteristics of SSL representations leading to high downstream accuracies.
method Characterized ideal properties and derived necessary and sufficient conditions.
result Improved SSL methods and derived new objectives for contrastive and non-contrastive learning.
Diffusion models generate private synthetic images with high quality.
problem Generating private synthetic images from sensitive datasets.
method Differentially private fine-tuning of pre-trained diffusion models.
result Improved synthetic data quality and accuracy compared to non-private models.
(Bolukbasi et al., 2016) demonstrated that pretrained word embeddings can inherit gender bias from the data they were trained on. We investigate how this bias affects downstream classification tasks, using the case study of occupation classification (De-Arteaga et al.,2019). We show that traditional techniques for debi…
Improved model accuracy can reduce overall user accuracy in competitive markets.
problem The impact of model competition on overall user accuracy.
method Defined a model of competition for classification tasks and used data representations to study the effect of scale.
result Improving data representation quality can decrease overall predictive accuracy across users (social welfare) in a competitive market.
BetaDataWeighter learns weights for unlabelled data to improve self-supervised learning accuracy.
problem Improving unsupervised representations with domain shift between unlabelled and target data.
method Learning Bayesian instance weights for unlabelled data to prioritize useful instances.
result BetaDataWeighter achieves highest average accuracy and prunes up to 78% of images without significant loss in accuracy.
DHOG improves unsupervised clustering accuracy on image benchmarks.
problem Local optima in mutual information maximisation lead to suboptimal representations.
method Deep hierarchical object grouping (DHOG) computes multiple discrete representations in a hierarchical order.
result DHOG achieves new state-of-the-art results on three main benchmarks.
Paper shows distillation can maintain low model churn.
problem Maintaining low model churn in real-world systems.
method Equivalence between distillation and explicit churn constraints.
result Distillation significantly reduces model churn with minimal accuracy loss.
Self-supervised method improves representation learning for better accuracy.
problem Improving representation learning without manual annotation.
method Proposes a novel self-supervised formulation of relational reasoning.
result Self-supervised relational reasoning outperforms state-of-the-art models by 14% in accuracy.
OneShotSTL efficiently decomposes time series online, improving speed and accuracy.
problem Real-time analysis of time series data with low processing delay.
method Online seasonal-trend decomposition algorithm with O(1) update time complexity.
result 1,000 times faster than batch methods with comparable accuracy.
Develops a method to ensure accuracy of few-shot transfer learning models.
problem Lack of generalization guarantees for low-data transfer learning.
method Trains a distribution over PEFT parameters using upstream tasks and samples plausible PEFTs for downstream tasks.
result Demonstrates non-vacuous generalization guarantees compared to existing methods in the low-shot regime.
Geospatial ML models need special evaluation methods due to their unique challenges.
problem Evaluating geospatial machine learning models is challenging due to their specific characteristics.
method Delineated unique challenges and proposed concrete takeaways for improving geospatial model evaluations.
result Concrete takeaways for improving evaluations of geospatial model performance.
Self-supervised learning improves few-shot classification and segmentation on point clouds.
problem Efficiently learn from limited labeled data in point cloud applications.
method Hierarchical cover-tree partitioning for self-supervised pre-training; restricted to support set for few-shot learning.
result Self-supervised learning significantly improves downstream classification and segmentation accuracy.
This paper improves cross-domain learning using random forests for manifold alignment.
problem Improving cross-domain learning and feature integration.
method Semi-supervised manifold alignment using random forest proximities.
result Random forest proximities enhance downstream classification accuracy.
Study shows uncertainty of deep learning models can be measured from their embeddings.
problem Uncertainty in contrastive learning models for critical applications.
method Estimating the distribution of training data in embedding space and accounting for local consistency.
result Uncertainty of an embedding vector correlates strongly with downstream accuracy.
In natural language processing, it has been observed recently that generalization could be greatly improved by finetuning a large-scale language model pretrained on a large unlabeled corpus. Despite its recent success and wide adoption, finetuning a large pretrained language model on a downstream task is prone to degen…
New method controls bias in training data for fair outcomes.
problem Ensuring equal treatment between different groups in machine learning.
method Contrastive information estimation to control mutual information between representations and protected attributes.
result Our method provides strong theoretical guarantees on the parity of any downstream algorithm.
New method uses imperfect LLM annotations for valid statistical inference in social science.
problem Inaccurate large language model annotations in social science research.
method Design-based supervised learning (DSL) combining imperfect LLM surrogates with gold-standard labels.
result DSL provides valid statistical inference with comparable predictive accuracy to existing methods.
Paper introduces a core-periphery model for identifying informative network structures.
problem Noise and bias in non-informative periphery structures obscure the informative core in complex networks.
method Spectral algorithms for core identification as a preprocessing step for network analysis.
result The proposed method outperforms traditional core-periphery methods in various downstream tasks.
Study finds significant instability in node embeddings due to randomness.
problem Stability of node embeddings under random variations.
method Evaluated five node embedding algorithms (HOPE, LINE, node2vec, SDNE, GraphSAGE) on synthetic and empirical graphs.
result Significant instability in embedding spaces and downstream task accuracy.
New theory explains contrastive learning via overlapping augmented views.
problem Lack of theoretical understanding of contrastive learning.
method Augmentation overlap perspective to improve downstream performance.
result Asymptotically closed bounds for downstream performance under weaker assumptions.
The paper tackles fair representation learning by smoothing feature mappings.
problem Legal liability for discriminatory use of data by organizations.
method Mapping features to a fair representation space, certifying fairness through chi-squared mutual information.
result Smoothing representation distribution provides generalization guarantees of fairness and maintains accuracy for downstream tasks.
Recently, pre-trained language representation flourishes as the mainstay of the natural language understanding community, e.g., BERT. These pre-trained language representations can create state-of-the-art results on a wide range of downstream tasks. Along with continuous significant performance improvement, the size an…
Networks are powerful data structures, but are challenging to work with for conventional machine learning methods. Network Embedding (NE) methods attempt to resolve this by learning vector representations for the nodes, for subsequent use in downstream machine learning tasks. Link Prediction (LP) is one such downstream…
Fine-grained pretraining improves neural network's ability to learn rare features.
problem Improving generalization in deep learning models.
method Introducing a hierarchical multi-view structure to confine input data distribution.
result Fine-grained pretraining leads to better accuracy on hard downstream test samples.
Adversarially trained transformers can learn robustly across tasks with minimal tuning.
problem Adversarial attacks and the high cost of adversarial training.
method Adversarial pretraining followed by in-context learning.
result Single-layer linear transformers can generalize robustly to unseen tasks.
This paper examines how labeling error affects contrastive learning and proposes data dimensionality reduction methods to mitigate its impact.
problem The impact of labeling error on the performance of contrastive learning.
method Data dimensionality reduction methods (e.g., SVD) are applied to reduce false positive samples and improve downstream classification accuracy.
result Data dimensionality reduction methods can mitigate the negative impacts of labeling error on downstream classification performance.
New approach optimizes decisions based on uncertainty in predictions.
problem Mismatch between prediction accuracy and decision loss in sequential design.
method Directional uncertainty-guided approach to sequential experimental design.
result Directional uncertainty-based design stops earlier and performs better.
Theoretical analysis shows pretext-based self-supervised learning can be boosted by downstream data under certain conditions.
problem Theoretical analysis of pretext-based self-supervised learning and downstream data refinement.
method Theoretical analysis and experiments on synthetic and real-world datasets.
result Theoretical lower bounds and experiments show that downstream data refinement can boost or hurt performance depending on conditions.
Object detection in point clouds is an important aspect of many robotics applications such as autonomous driving. In this paper we consider the problem of encoding a point cloud into a format appropriate for a downstream detection pipeline. Recent literature suggests two types of encoders; fixed encoders tend to be fas…
Ensemble pre-trained models for low data transfer learning.
problem Training good models from scratch in low data regime.
method Fine-tune nearest-neighbour ranked pre-trained models to create ensembles.
result Achieves state-of-the-art performance with lower inference budget.
Transfer learning using deep neural networks as feature extractors has become increasingly popular over the past few years. It allows to obtain state-of-the-art accuracy on datasets too small to train a deep neural network on its own, and it provides cutting edge descriptors that, combined with nonparametric learning m…
MPP trains a transformer to predict multiple physical systems, improving accuracy across various tasks.
problem Training models for specific physical systems is inefficient and requires fine-tuning.
method MPP trains a shared transformer on multiple heterogeneous physical systems, projecting fields into a shared embedding space.
result A single MPP-pretrained transformer outperforms task-specific models on all pretraining sub-tasks and downstream tasks.
BCDP enhances privacy by protecting sensitive features more precisely.
problem Uniform privacy protection in LDP degrades performance for sensitive features.
method Bayesian Coordinate Differential Privacy (BCDP) adjusts privacy protection per feature sensitivity.
result BCDP improves accuracy in downstream tasks without sacrificing privacy.
Paper defines and solves a problem in representation learning to ensure fairness with high confidence.
problem Learning fair representations with high confidence guarantees for all downstream tasks.
method Formally defines the problem, introduces FRG framework, proves high probability fairness, and demonstrates effectiveness empirically.
result FRG framework provides high-confidence guarantees for limiting unfairness across all downstream models and tasks.
Few-shot DP image classification models need more data as privacy increases.
problem Few-shot DP image classification challenges in personalization and federated learning.
method Exhaustive experiments on various parameters affecting few-shot DP accuracy and vulnerability.
result Increasing shots per class is necessary for DP accuracy as privacy increases.
New findings show BERT subnetworks can train independently and transfer to various tasks.
problem Finding smaller subnetworks that can train independently and transfer to other tasks.
method Examined pre-trained BERT models for subnetworks that can train independently and transfer to various downstream tasks.
result Found subnetworks at 40% to 90% sparsity that can train independently and transfer to various tasks.