Generative models need per-sample confidence scores to improve quality and stability.
problem Generative models produce unreliable outputs and lack confidence measures.
method Flow Matching with Confidence (FMwC) injects noise and integrates it through the network, providing per-sample confidence scores.
result The confidence score correlates with the velocity field's divergence, offering insights into generative processes.
New method estimates grouping loss in neural networks to improve confidence scores.
problem Improving confidence scores in neural networks to reflect true posterior probabilities.
method Proposed an estimator to approximate the grouping loss.
result Modern neural networks exhibit grouping loss, especially in distribution shifts.
We propose a novel confidence scoring mechanism for deep neural networks based on a two-model paradigm involving a base model and a meta-model. The confidence score is learned by the meta-model observing the base model succeeding/failing at its task. As features to the meta-model, we investigate linear classifier probe…
The paper introduces multicalibration to improve confidence scores in LLMs.
problem Improving the reliability and interpretability of confidence scores for LLMs.
method Forming groupings of prompt/completion pairs correlated with correctness, using clustering and self-annotation. Developing multicalibration algorithms to reduce overfitting.
result Our techniques yield confidence scores that significantly improve calibration and accuracy compared to existing methods.
New method reduces confidence interval sizes for causal inference.
problem Inaccurate propensity scores and extreme scores cause large confidence intervals.
method Data-dependent Coarse IPW (CIPW) estimators.
result Robust CIPW estimators reduce confidence interval sizes to ε+1/√n.
We present a new method for computing ASR word confidences that effectively mitigates the effect of ASR errors for diverse downstream applications, improves the word error rate of the 1-best result, and allows better comparison of scores across different models. We propose 1) a new method for modeling word confidence u…
Develops spatial uncertainty guarantees for image segmentation models.
problem Ensuring reliable segmentation predictions for biomedical images.
method Adapting conformal inference to imaging, using transformed logit scores and calibration datasets.
result Confidence sets provide spatial uncertainty guarantees with desired probability.
TeLeS improves ASR confidence estimation by considering temporal alignment and lexical errors.
problem Inaccurate confidence scores from E2E ASR models, especially for overconfident predictions.
method Proposes TeLeS, a novel confidence score that considers temporal alignment and lexical errors, and uses shrinkage loss to handle data imbalance.
result TeLeS generalizes well across different languages and ASR models, leading to significant WER reduction.
We propose a generic framework to calibrate accuracy and confidence of a prediction in deep neural networks through stochastic inferences. We interpret stochastic regularization using a Bayesian model, and analyze the relation between predictive uncertainty of networks and variance of the prediction scores obtained by …
SMURF-THP improves Transformer Hawkes process models by providing uncertainty quantification.
problem Uncertainty quantification for Transformer Hawkes process predictions.
method Score matching for learning the score function of event arrival times.
result SMURF-THP outperforms likelihood-based methods in confidence calibration.
CCE improves anomaly detection metrics by measuring both confidence and consistency.
problem Existing anomaly detection metrics lack discriminative power, hyperparameter dependency, and robustness to perturbations.
method CCE uses Bayesian estimation to quantify uncertainty and constructs global and event-level confidence and consistency scores.
result CCE demonstrates strict boundedness, robustness, and linear time complexity.
Predicting properties of nodes in a graph is an important problem with applications in a variety of domains. Graph-based Semi-Supervised Learning (SSL) methods aim to address this problem by labeling a small subset of the nodes as seeds and then utilizing the graph structure to predict label scores for the rest of the …
The reliable measurement of confidence in classifiers' predictions is very important for many applications and is, therefore, an important part of classifier design. Yet, although deep learning has received tremendous attention in recent years, not much progress has been made in quantifying the prediction confidence of…
Imitation learning (IL) aims to learn an optimal policy from demonstrations. However, such demonstrations are often imperfect since collecting optimal ones is costly. To effectively learn from imperfect demonstrations, we propose a novel approach that utilizes confidence scores, which describe the quality of demonstrat…
In learning with noisy labels, for every instance, its label can randomly walk to other classes following a transition distribution which is named a noise model. Well-studied noise models are all instance-independent, namely, the transition depends only on the original label but not the instance itself, and thus they a…
Detecting depression early from social media texts.
problem Early diagnosis and prevention of depression.
method Topic analysis and learned confidence scores.
result Achieved good results compared to state of the art.
New method improves certified robustness for classifier confidence.
problem Certifying confidence in classifier predictions.
method Randomized smoothing with modified Neyman-Pearson lemma.
result Certified radii for prediction confidence improved.
BoC probe assesses neural network confidence coherence, revealing architecture-specific uncertainty.
problem Poor calibration and OOD detection in neural networks.
method Bag-of-Coins (BoC) probe compares softmax confidence to pairwise dominance probabilities.
result BoC reveals clear ID/OOD separation for some architectures but not others.
The paper improves ranking by integrating covariates and sparse intrinsic scores.
problem Ranking items with incomplete preference scores explained by covariates.
method Extends BTL model with covariate information and sparse intrinsic scores, using penalized MLE.
result Developed debiased estimator for penalized MLE with distributional properties.
The study examines when to trust confidence thresholding in pseudo-labelling regression.
problem Calibrated probabilities from classifiers used for pseudo-labelling need careful handling to avoid bias in downstream regression.
method Developed a diagnostic apparatus to predict and bound the bias induced by confidence thresholding, derived a closed-form expression for the attenuation bias.
result The bias can be predicted from the residual score variance V∗, motivating a structural separation between classifier features and downstream controls. Cost-effective method detects language model hallucinations.
problem Detecting unreliable outputs from LLMs.
method Pipeline including confidence score, input attributes calibration, and thresholding.
result Multi-scoring framework outperforms individual methods and reduces computational cost.
WS-KDE provides robust confidence bounds for stochastic functions.
problem Optimizing time-consuming black-box functions with stochastic outputs.
method Wilson Score Kernel Density Estimation (WS-KDE) for Bayesian optimization.
result WS-KDE provides reliable confidence bounds for any stochastic function.
Estimating machine learning performance 'in the wild' is both an important and unsolved problem. In this paper, we seek to examine, understand, and predict the pointwise competence of classification models. Our contributions are twofold: First, we establish a statistically rigorous definition of competence that general…
New method improves auto-labeling accuracy by optimizing confidence functions.
problem Overconfident model scores lead to poor TBAL performance.
method Developed a new post-hoc method, Colander, to optimize TBAL confidence functions.
result Achieves up to 60% improvement in coverage over baseline methods.
Study finds simple model-agreement scores perform well in various error estimation scenarios.
problem Evaluating model performance on unseen distributions using disparate scoring functions.
method Rigorously studied popular scoring functions (confidence, local manifold smoothness, model agreement) independently of mechanism choice.
result Simple model-agreement scores outperform confidence- and smoothness-based scores in realistic settings with compromised training data.
A new OOD detector using an overlap index improves accuracy without high computational costs.
problem Effective OOD detection for machine learning models in open-world scenarios.
method Proposes an overlap index-based confidence score function for OOD detection.
result The proposed method achieves competitive accuracy with lower computational costs compared to state-of-the-art detectors.
Study three types of uncertainty quantification for binary classification without distributional assumptions.
problem Uncertainty quantification for binary classification in a distribution-free setting.
method Established theorems connecting calibration, confidence intervals, and prediction sets for score-based classifiers.
result Distribution-free calibration is only possible using scoring functions that partition feature space into countably many sets.
The Mahalanobis distance-based confidence score, a recently proposed anomaly detection method for pre-trained neural classifiers, achieves state-of-the-art performance on both out-of-distribution (OoD) and adversarial examples detection. This work analyzes why this method exhibits such strong performance in practical s…
Paper compares two forecasters using novel online inference methods.
problem How to compare forecasters without distributional assumptions.
method Confidence sequences and game-theoretic statistical framework for sequential testing.
result Valid methods for comparing forecasters without distributional assumptions.
New confidence intervals improve treatment effect estimation in randomized experiments.
problem Improving confidence intervals for treatment effects in randomized experiments.
method Systematic exploitation of negative dependence or variance adaptivity.
result Achieved nonasymptotic confidence intervals with the same effective sample size as asymptotic ones.
Knowing when a classifier's prediction can be trusted is useful in many applications and critical for safely using AI. While the bulk of the effort in machine learning research has been towards improving classifier performance, understanding when a classifier's predictions should and should not be trusted has received …
New algorithm for precise changepoint localization without assumptions.
problem Offline changepoint localization in arbitrary distributions.
method Distribution-free algorithm CONformal CHangepoint localization (CONCH) using exchangeability arguments.
result Derives principled score functions for informative and small confidence sets with normalized length shrinking to zero.
Focal loss improves classification but not class-posterior probability estimation.
problem Improving class-posterior probability estimation from focal loss.
method Proved classification-calibration and derived a transformation to recover true class-posterior probabilities.
result A transformation of the confidence score from focal loss minimization allows recovery of true class-posterior probabilities.
We consider the problem of uncertainty assessment for low dimensional components in high dimensional models. Specifically, we propose a decorrelated score function to handle the impact of high dimensional nuisance parameters. We consider both hypothesis tests and confidence regions for generic penalized M-estimators. U…
Introduces privilege scores to measure and interpret protected attribute-related privilege in machine learning models.
problem Lack of explicit formulation of non-neutrality in fairness-aware machine learning methods.
method Privilege scores (PS) and privilege score contributions (PSCs) to measure and interpret protected attribute-related privilege.
result Demonstrates the broad applicability of PS and PSCs in gender and racial privilege in mortgage and college admissions applications.
GPT-4 assesses its confidence in answering USMLE questions with and without feedback.
problem Understanding AI's performance in healthcare applications, especially in sensitive areas like medical education.
method Used a prompting technique to evaluate GPT-4's confidence scores before and after answering USMLE questions, categorized into with and without feedback.
result Feedback influences relative confidence but doesn't consistently increase or decrease it.
New attacks can infer model training membership using only label predictions, not confidence.
problem Inferring whether a data point was used to train a machine learning model.
method Evaluate model's predicted labels under perturbations to infer membership.
result Label-only attacks perform as well as confidence-based attacks and break defenses that rely on confidence masking.
One of the problems on the way to successful implementation of neural networks is the quality of annotation. For instance, different annotators can annotate images in a different way and very often their decisions do not match exactly and in extreme cases are even mutually exclusive which results in noisy annotations a…
Semantic Embeddings are a popular way to represent knowledge in the field of zero-shot learning. We observe their interpretability and discuss their potential utility in a safety-critical context. Concretely, we propose to use them to add introspection and error detection capabilities to neural network classifiers. Fir…
Improved conformal prediction for better conditional coverage of classifier predictions.
problem Achieving exact conditional coverage in finite samples for prediction sets.
method Developed a variant of conformal prediction targeting coverage conditional on confidence and trust score.
result Empirically improved conditional coverage properties compared to standard conformal prediction.
This work introduces COLA, a strategy to aggregate conformal prediction sets efficiently.
problem Efficiently combining multiple conformity scores to reduce prediction set size.
method Introduces COnfidence-Level Allocation (COLA) to optimally allocate confidence levels across sets.
result COLA achieves smaller prediction sets than state-of-the-art methods while maintaining valid coverage.
The paper improves confidence set construction for statistical inference.
problem Constructing reliable confidence sets in statistical inference.
method Establishes a finite-sample bound using effective dimension and generalized self-concordance.
result Developed a confidence set adapted to optimization landscapes.
This paper proposes a decorrelation-based approach to test hypotheses and construct confidence intervals for the low dimensional component of high dimensional proportional hazards models. Motivated by the geometric projection principle, we propose new decorrelated score, Wald and partial likelihood ratio statistics. Wi…
Confidence-based filtering reveals latent structure in diffusion models.
problem Unclear latent structure in diffusion models.
method Confidence scores from a classifier.
result Class-relevant latent structure emerges under confidence-based filtering.
New method improves conditional coverage of conformal prediction.
problem Improving conditional coverage in conformal prediction.
method Trainable transformation of conformity scores to improve conditional coverage.
result Highly adaptive to local data structure, outperforming existing methods.
In this paper, we propose a new method called ProfWeight for transferring information from a pre-trained deep neural network that has a high test accuracy to a simpler interpretable model or a very shallow network of low complexity and a priori low test accuracy. We are motivated by applications in interpretability and…
This work introduces a bias-variance decomposition for proper scores, improving uncertainty estimation in predictive models.
problem Reliable uncertainty estimation for predictions in safety-critical applications, especially under domain drift.
method Developed a general bias-variance decomposition for proper scores, introducing the Bregman Information as the variance term.
result The decomposition provides novel formulations for different predictive tasks, including classification and model ensembles.
Certified calibration methods protect model confidence from adversarial attacks.
problem Adversarial attacks degrade model calibration, reducing confidence in predictions.
method Developed certified calibration methods to provide worst-case bounds on calibration under adversarial perturbations.
result Certified calibration methods produce analytic and approximate bounds for the Brier score and expected calibration error.