Tag
This article benchmarks various blackbox and whitebox methods for LLM confidence estimation, including verbalized confidence, linguistic uncertainty, reasoning-length, P(Answer), P(True), and self-consistency, comparing their effectiveness for tasks like active learning and safety classification.
This paper presents BattVAE-GP, a hybrid physics-probabilistic framework that combines a Variational Autoencoder with a sparse multitask Gaussian Process to generate and interpolate long-horizon battery degradation trajectories with uncertainty estimates, enabling efficient surrogate modeling for lithium-ion battery health prediction.
Introduces PhiCalNet, a neural network for single-shot fringe projection profilometry that avoids shape-prior shortcuts by outputting wrapped phase and using a fixed calibration layer, achieving 3.3x lower error than baseline UNet on a synthetic benchmark.
The paper proposes a lightweight method that reformulates regression-based INR training as a classification task by discretizing continuous targets into bins, enabling flexible distribution modeling for error-aware uncertainty estimation in scientific data compression.
Proposes learned predictive ambiguity sets (LPAS) for distributionally robust optimization, where a deep contextual model outputs a nominal scenario distribution, state-dependent Wasserstein radius, and ground metric, trained with decision loss and calibration. Applied to portfolio optimization on S&P 500 data, the method achieves higher returns and Sharpe ratio with reduced conservatism compared to fixed-radius baselines.
This paper empirically characterizes uncertainty in thinking-mode visual language models, demonstrating that the thinking chain entropy is a more reliable signal for hallucination detection than conventional answer token distribution, which collapses in these models.
This paper introduces GraphEVAL, a graph-based framework for quantifying uncertainty in LLM reasoning, and proposes a new metric, Graph Reasoning Coherence Score (GRCS), that captures semantic-structural consensus and detects confident hallucinations. The authors also present Graph Self-Consistency (GSC), a decoding strategy that prioritizes reasoning fidelity over nominal accuracy.
This paper studies which signals best predict correctness in text-to-SQL for selective prediction. It finds that verification-based signals from LLM judges outperform black-box statistical signals like self-consistency, and that a two-provider ensemble achieves 0.82 AUROC with well-calibrated probabilities.
This paper reveals that standard marginal conformal prediction fails to cover minority classes in imbalanced virtual screening datasets, and demonstrates that class-conditional (Mondrian) conformal prediction restores per-class coverage.
This paper introduces Heckman-corrected epistemic uncertainty to address selection on unobservables in machine learning, demonstrating that importance weighting fails when selection depends on unobservables correlated with outcomes. The method restores calibration in controlled experiments and real data, outperforming standard UQ baselines.
SpanUQ introduces a lightweight probe for span-level uncertainty quantification in LLMs, using a DETR-style decoder and mixture of Beta distributions, achieving superior error localization and faster inference than sampling-based methods.
This paper introduces Agentic SABRE, an uncertainty-aware neuro-symbolic multi-agent framework for adaptive ransomware detection that fuses semantic and behavioral evidence with Monte Carlo Dropout and interpretable risk-uncertainty triage, demonstrating improved robustness and explainability.
Introduces TRIE, an evaluation framework for stochastic PDE surrogates that tests reproduction of invariant measures, trustworthy predictive uncertainty, and efficiency. Benchmarks pointwise-trained neural surrogates, approximate uncertainty methods, and generative models on two SPDEs, finding generative models most consistent.
Patch-PODiff-ViT introduces a structured latent diffusion framework using patchwise Proper Orthogonal Decomposition (POD) for super-resolution and uncertainty quantification, enabling efficient diffusion with a fixed linear orthonormal basis and analytic propagation of predictive variance.
This paper presents a sparse Gaussian process framework for quantile regression that uses a Laplace approximation for posterior inference and variance-based mechanisms for adaptive inducing-input placement and data acquisition.
This paper presents a transferable learned temporal prior for outbreak transmission reconstruction, demonstrating improved performance on a real Andes virus benchmark and highlighting the importance of quantifying uncertainty in transmission labels.
This paper introduces the NHANES Accelerometry Cardiometabolic Benchmark, a population-representative tabular dataset for predicting cardiometabolic risk from accelerometry data, and evaluates ridge regression, XGBoost, and TabPFN v2 with uncertainty quantification using conformal prediction.
This paper characterizes when conformal risk control can certify structured LLM outputs, proving impossibility bounds and analyzing certification hierarchies across different bounds. Empirical validation on six open-weight models shows that hard configurations are uncertifiable at low risk levels but practical certification is achievable at relaxed targets.
Introduces a conformal prediction method for ordinal classification using the ranked probability score as a nonconformity function, producing median-centered contiguous prediction sets and achieving favorable balance between set width and ordinal miscoverage.
This paper formulates orchestration of coding agents as cost-sensitive sequential hypothesis testing using a Bayesian controller that dynamically decides when to gather evidence, refine, verify, or stop. Experiments across six generators and nine benchmarks show Bayesian control is most valuable when verification is costly and critics are informative but imperfect.