RealCause provides a realistic benchmark for causal inference.
problem Lack of a reliable benchmark for comparing causal effect estimators.
method Flexible generative models to create a benchmark that is both ground-truth and realistic.
result Evaluation of over 1500 causal estimators provides evidence for choosing hyperparameters using predictive metrics.
Universally valid ground truth is almost impossible to obtain or would come at a very high cost. For supervised learning without universally valid ground truth, a recommended approach is applying crowdsourcing: Gathering a large data set annotated by multiple individuals of varying possibly expertise levels and inferri…
Ranking a set of objects involves establishing an order allowing for comparisons between any pair of objects in the set. Oftentimes, due to the unavailability of a ground truth of ranked orders, researchers resort to obtaining judgments from multiple annotators followed by inferring the ground truth based on the collec…
Error correction for correlated regressors using compressed sensing.
problem Measuring errors in regression tasks with sparse correlations.
method Compressed sensing applied to error signals, with ℓ1-minimization for uniqueness. result Correct solution recovery in settings of error correction.
Bayesian model improves truth inference from highly redundant crowd annotations.
problem Inferring true annotations from highly redundant crowd annotations.
method Bayesian graphical model with conjugate priors and iterative expectation-maximisation inference.
result Our technique significantly outperforms majority vote heuristic at one-sided level 0.025.
An unsupervised neural network learns event truths from social network data.
problem Estimating event truths from conflicting opinions in social networks.
method Autoencoder learns relationships, Bayesian network models agent reliability and social relationships, variational inference estimates hidden variables and parameters.
result The approach outperforms state-of-the-art methods on real datasets.
We investigate the problem of truth discovery based on opinions from multiple agents who may be unreliable or biased. We consider the case where agents' reliabilities or biases are correlated if they belong to the same community, which defines a group of agents with similar opinions regarding a particular event. An age…
Estimates classifier errors without ground truth using algebraic geometry.
problem Lack of ground truth in real-world production systems.
method Non-parametric estimation using algebraic geometry to solve the self-assessment problem.
result Accuracy estimators are better than one part in a hundred.
In the modern era, abundant information is easily accessible from various sources, however only a few of these sources are reliable as they mostly contain unverified contents. We develop a system to validate the truthfulness of a given statement together with underlying evidence. The proposed system provides supporting…
Polymarket-v1 Database tracks 1.2B trades across 1.3M markets with 100% ground-truth direction.
problem Lack of ground-truth data in prediction markets archives.
method Ground-truth archive of 1.2B trades from Polymarket's CTF Exchange.
result Ground-truth data reveals systematic errors in microstructure metrics.
Improves NMT by sampling context from predicted sequence during training.
problem Error accumulation and overcorrection in NMT due to mismatched training and inference contexts.
method Samples context words from both ground truth and predicted sequences during training.
result Significant improvements on multiple datasets, including Chinese->English and WMT'14 English->German.
Latent truth discovery, LTD for short, refers to the problem of aggregating ltiple claims from various sources in order to estimate the plausibility of atements about entities. In the absence of a ground truth, this problem is highly challenging, when some sources provide conflicting claims and others no claims at all.…
PPI uses proxy data to improve inference from limited labels across related tasks.
problem Statistical inference with limited labels across multiple related tasks.
method Prediction-powered inference framework that uses cross-task recalibration to improve power and accuracy.
result Cross-task recalibration can substantially reduce confidence interval widths when labels are scarce.
A method for inferring ground-truth signals from degraded sensor data.
problem Inferring ground-truth signals from multiple degraded sensor signals.
method Iterative correction of degraded signals using a Bayesian multi-sensor data fusion method.
result The method effectively infers ground-truth signals from noisy and degraded sensor data.
Misspecification-Aware Simulation-Based Inference via Side-Channel Guidance
problem Simulation-based inference (SBI) of latent parameters is hindered by simulator misspecification.
method Misspecification-Aware Simulation-Based Inference (MA-SBI) turns side-channel text into a posterior correction.
result MA-SBI matches the oracle posterior across 10 seeds and two backbones.
RoPE framework calibrates misspecified simulators for reliable inference.
problem Misspecification compromises reliability of simulation-based inference.
method Data-driven calibration using optimal transport and a small calibration set.
result RoPE framework improves inference accuracy and uncertainty calibration.
New metric improves latent dynamics inference from neural data.
problem Limitations of co-smoothing in predicting latent dynamics.
method Few-shot co-smoothing to assess latent dynamics.
result High co-smoothing models often have extraneous dynamics, which few-shot co-smoothing detects.
Self-supervised methods learn from noisy data alone, useful for imaging problems.
problem Inferring signals from noisy and incomplete observations.
method Learning a solver from measurement data alone, without ground-truth references.
result Self-supervised methods can learn meaningful estimates from noisy data.
New method falsifies causal discovery results without ground truth.
problem Evaluation of causal discovery algorithms without ground truth data.
method Detects incompatibilities between causal graphs learned on different subsets of variables.
result Detection of incompatibilities can falsify wrongly inferred causal relations.
Improves spatio-temporal forecasting by reducing errors between training and inference.
problem Accumulation of small errors in Seq2Seq models during inference due to different distributions of training and inference phases.
method Curriculum learning based on Temporal Progressive Growing Sampling to replace some ground-truth context with generated predictions.
result Better models long-term dependencies and outperforms baseline approaches on two datasets.
Paper recovers top-two answers and confusion probability in multi-choice crowdsourcing.
problem Recovering top-two answers and confusion probability in multi-choice crowdsourcing tasks.
method Proposes a two-stage inference algorithm based on a model quantifying task difficulty and worker reliability.
result Achieves minimax optimal convergence rate and outperforms other algorithms in synthetic and real data experiments.
New method accounts for uncertainty in medical AI evaluations.
problem Uncertainty in ground truth affects AI model performance estimates.
method Statistical aggregation approach to infer probabilities of medical conditions.
result Performance estimates are significantly lower when uncertainty is accounted for.
New method selects best HTE estimator without ground-truth treatment effects.
problem Selecting best HTE estimator from multiple candidates.
method Cross-fitted, exponentially weighted test statistic with two-way sample splitting.
result Empirically, reliable error control and reduced false selections.
A popular approach for large scale data annotation tasks is crowdsourcing, wherein each data point is labeled by multiple noisy annotators. We consider the problem of inferring ground truth from noisy ordinal labels obtained from multiple annotators of varying and unknown expertise levels. Annotation models for ordinal…
The presence of a sparse "truth" has been a constant assumption in the theoretical analysis of sparse PCA and is often implicit in its methodological development. This naturally raises questions about the properties of sparse PCA methods and how they depend on the assumption of sparsity. Under what conditions can the r…
Large Language Models (LLMs) have demonstrated impressive generalization capabilities across various tasks, but their claim to practical relevance is still mired by concerns on their reliability. Recent works have proposed examining the activations produced by an LLM at inference time to assess whether its answer to a …
Obtaining enough labeled data to robustly train complex discriminative models is a major bottleneck in the machine learning pipeline. A popular solution is combining multiple sources of weak supervision using generative models. The structure of these models affects training label quality, but is difficult to learn with…
Improved VB algorithm for high-dimensional logistic regression with theoretical guarantees.
problem Sparse high-dimensional logistic regression model selection.
method Spike and slab variational Bayes approximation.
result Optimal convergence rates in ℓ2 and prediction loss for sparse truths. Causal inference from observational data often assumes "ignorability," that all confounders are observed. This assumption is standard yet untestable. However, many scientific studies involve multiple causes, different variables whose effects are simultaneously of interest. We propose the deconfounder, an algorithm that…
Empirical study shows overparameterization benefits unsupervised learning of latent variable models.
problem Improving optimization landscape in unsupervised learning with overparameterization.
method Synthetic and semi-synthetic experiments with various models and training algorithms.
result Overparameterization significantly increases the number of ground truth latent variables recovered.
Inference-Time Scaling can be extended to domains prone to systematic failure using intrinsic statistics.
problem Scaling inference time in domains prone to systematic failure
method Intrinsic Selection (iS), Intrinsic Particle Filtering (iPF), and Particle Distillation (dPF)
result Intrinsic Selection improves engineering design selection by 20% and pass@1 by 6.1 points on average.
New truthful calibration errors improve model ranking in multiclass prediction.
problem Non-truthful calibration errors can mislead model comparisons.
method Introduced perfectly truthful calibration errors for multiclass predictions.
result Truthful calibration errors preserve decision-theoretic dominance and stabilize model rankings.
This paper introduces NPR, a technique to improve Bayesian inference for multi-modal, high-dimensional simulations.
problem Challenges in Bayesian inference for multi-modal, high-dimensional simulations.
method Introduces Neural Posterior Regularization (NPR) to enforce exploration of input parameter space.
result Empirically validated that NPR significantly improves performance on various simulation tasks.
Paper presents a machine learning method to improve significance tests for misspecified linear models.
problem Misspecification of linear assumptions in social science models leads to inaccurate significance levels.
method Apply machine learning to fit ground truth function, calculate linear approximation, and adjust the estimator.
result The method significantly outperforms linear regression for non-linear ground truth functions.
Crowdsourcing has been proven to be an effective and efficient tool to annotate large datasets. User annotations are often noisy, so methods to combine the annotations to produce reliable estimates of the ground truth are necessary. We claim that considering the existence of clusters of users in this combination step c…
DBLE improves confidence calibration of DNNs by learning distances in representation space.
problem Poor confidence calibration of deep neural networks (DNNs).
method DBLE trains a confidence model jointly with the classification model, using distances in the representation space.
result DBLE outperforms alternative single-model confidence calibration approaches and ensemble methods.
This work improves graph inference using the degree-4 sum-of-squares hierarchy.
problem Recovering ground-truth binary labelings from corrupted edge observations.
method Apply the degree-4 sum-of-squares hierarchy to a quadratic combinatorial optimization problem.
result The solution of the dual problem is related to edge weights of Johnson and Kneser graphs.
Eye Movement analysis with Hidden Markov Models (EMHMM) is a method for modeling eye fixation sequences using hidden Markov models (HMMs). In this report, we run a simulation study to investigate the estimation error for learning HMMs with variational Bayesian inference, with respect to the number of sequences and the …
New calibration measure SSCE ensures truthful prediction, unlike existing measures.
problem Ensuring truthful calibration measures in sequential prediction.
method Introduced a new calibration measure, Subsampled Smooth Calibration Error (SSCE).
result SSCE ensures truthful prediction, while existing measures are far from truthful.
A new perfectly truthful calibration measure improves prediction reliability.
problem Improving the reliability of predictions by ensuring they are conditionally unbiased.
method Designing a simple, perfectly truthful calibration measure called ATB.
result ATB is the first perfectly truthful calibration measure in the batch setting.
Efficiently estimates marginal posteriors for complex simulations.
problem Bayesian inference in high-dimensional, intractable likelihood scenarios.
method Simulates and estimates low-dimensional marginal posteriors, using truncated indicators.
result Simulator efficiency and robustness testing of inference results.
The unprecedented demand for large amount of data has catalyzed the trend of combining human insights with machine learning techniques, which facilitate the use of crowdsourcing to enlist label information both effectively and efficiently. The classic work on crowdsourcing mainly focuses on the label inference problem …
New methods reduce bias in machine learning predictions for causal inference without extra data.
problem Machine learning predictions from satellite data shrink toward the mean, leading to biased causal estimates.
method Two post-hoc correction methods: Linear Calibration Correction (LCC) and Tweedie's approach, reduce shrinkage-induced bias.
result Tweedie's method yields nearly unbiased treatment-effect estimates, enabling multiple trials with a single map.
This paper refines human labeling as a measurement process, revealing four sources of variation.
problem Systematic variation in human labeling obscures model learning.
method Introduces a statistical framework to decompose labeling outcomes.
result Empirical evidence for four components of labeling variation.
SBIC improves real-time online binary classification from crowdsourced data.
problem Inferring ground truth from noisy crowdsourced data.
method Streaming Bayesian Inference for Crowdsourcing (SBIC).
result SBIC has low complexity, high accuracy, and provable guarantees.
In many security and healthcare systems, the detection and diagnosis systems use a sequence of sensors/tests. Each test outputs a prediction of the latent state and carries an inherent cost. However, the correctness of the predictions cannot be evaluated since the ground truth annotations may not be available. Our obje…
New causal distances improve evaluation of causal discovery algorithms.
problem Evaluating causal discovery algorithms using graphical distances is limited.
method Defined causal distances based on causal distributions rather than graphical structure.
result Improved evaluation of causal discovery algorithms on synthetic and real-world datasets.
Poisoning datasets can reveal private details of other users' training points.
problem Integrity and privacy of machine learning training data.
method Active inference attacks that poison a small fraction of the training dataset.
result Poisoning as little as 0.1% of the training dataset can significantly boost inference attacks.