Neural Information Processing Systems http://nips.cc/

An increasingly common setting in machine learning involves multiple parties, each with their own data, who want to jointly make predictions on future test points. Agents wish to benefit from the collective expertise of the full set of agents to make better predictions than they would individually, but may not be willing to release labeled data or model parameters.

Trustworthiness has become a key requirement for the deployment of artificial intelligence systems in safety-critical applications. Conventional evaluation metrics, such as accuracy and precision, fail to appropriately capture uncertainty or the reliability of model predictions, particularly under adversarial or degraded conditions. This paper introduces the Parallel Trust Assessment System (PaTAS), a framework for modeling and propagating trust in neural networks using Subjective Logic (SL). PaTAS operates in parallel with standard neural computation through Trust Nodes and Trust Functions that propagate input, parameter, and activation trust across the network. The framework defines a Parameter Trust Update mechanism to refine parameter reliability during training and an Inference-Path Trust Assessment (IPTA) method to compute instance-specific trust at inference. Experiments on real-world and adversarial datasets demonstrate that PaTAS produces interpretable, symmetric, and convergent trust estimates that complement accuracy and expose reliability gaps in poisoned, biased, or uncertain data scenarios. The results show that PaTAS effectively distinguishes between benign and adversarial inputs and identifies cases where model confidence diverges from actual reliability. By enabling transparent and quantifiable trust reasoning within neural architectures, PaTAS provides a foundation for evaluating model reliability across the AI lifecycle.

Knowing when a classifier's prediction can be trusted is useful in many applications

Mislabeled data is a pervasive issue that undermines the performance of machine learning systems in real-world applications. An effective approach to mitigate this problem is to detect mislabeled instances and subject them to special treatment, such as filtering or relabeling. Automatic mislabeling detection methods typically rely on training a base machine learning model and then probing it for each instance to obtain a trust score that each provided label is genuine or incorrect. The properties of this base model are thus of paramount importance. In this paper, we investigate the impact of calibrating this model. Our empirical results show that using calibration methods improves the accuracy and robustness of mislabeled instance detection, providing a practical and effective solution for industrial applications.

Mobile robots with some degree of autonomy could deliver significant advantages in high-risk missions such as search and rescue and firefighting. Integrated into a human-robot team (HRT), robots could work effectively to help search hazardous buildings. User trust is a key enabler for HRT, but during a mission, trust can be damaged. With distributed situation awareness, such as when team members are working in different locations, users may be inclined to doubt a robot's integrity if it declines to immediately change its priorities on request. In this paper, we present the results of a computer-based study investigating on-mission trust dynamics in a high-stakes human-robot teaming scenario. Participants (n = 38) played an interactive firefighting game alongside a robot teammate, where a trust violation occurs owing to the robot declining to help the user immediately. We find that when the robot provides an explanation for declining to help, trust better recovers over time, albeit following an initial drop that is comparable to a baseline condition where an explanation for refusal is not provided. Our findings indicate that trust can vary significantly during a mission, notably when robots do not immediately respond to user requests, but that this trust violation can be largely ameliorated over time if adequate explanation is provided.

Federated learning is vulnerable to poisoning and backdoor attacks under partial observability. We formulate defence as a partially observable sequential decision problem and introduce a trust-aware Deep Q-Network that integrates multi-signal evidence into client trust updates while optimizing a long-horizon robustness--accuracy objective. On CIFAR-10, we (i) establish a baseline showing steadily improving accuracy, (ii) show through a Dirichlet sweep that increased client overlap consistently improves accuracy and reduces ASR with stable detection, and (iii) demonstrate in a signal-budget study that accuracy remains steady while ASR increases and ROC-AUC declines as observability is reduced, which highlights that sequential belief updates mitigate weaker signals. Finally, a comparison with random, linear-Q, and policy gradient controllers confirms that DQN achieves the best robustness--accuracy trade-off.

Abstract--In machine learning, uncertainty quantification helps assess the reliability of model predictions, which is important in high-stakes scenarios. Traditional approaches often emphasize predictive accuracy, but there is a growing focus on incorporating uncertainty measures. While current methods often rely on quantile regression or Monte Carlo techniques, we propose a new approach using naturally occurring test sets and similarity measures (proximities) typically viewed as byproducts of random forests. Specifically, we form localized distributions of OOB errors around nearby points, defined using the proximities, to create prediction intervals for regression and trust scores for classification. By varying the number of nearby points, our intervals can be adjusted to achieve the desired coverage while retaining the flexibility that reflects the certainty of individual predictions. For classification, excluding points identified as unclassifiable by our method generally enhances the accuracy of the model and provides higher accuracy-rejection AUC scores than competing methods. Although traditional machine learning models usually provide point estimates, there is growing recognition of the need to incorporate uncertainty to support more informed decisions [1]. By quantifying uncertainty, users can assess the reliability of model outputs and better interpret results, especially for out-of-distribution samples through calibrated confidence estimates.

Aligning AI with human values is a pressing unsolved problem. To address the lack of quantitative metrics for value alignment, we propose EigenBench: a black-box method for comparatively benchmarking language models' values. Given an ensemble of models, a constitution describing a value system, and a dataset of scenarios, our method returns a vector of scores quantifying each model's alignment to the given constitution. To produce these scores, each model judges the outputs of other models across many scenarios, and these judgments are aggregated with EigenTrust (Kamvar et al., 2003), yielding scores that reflect a weighted consensus judgment of the whole ensemble. EigenBench uses no ground truth labels, as it is designed to quantify subjective traits for which reasonable judges may disagree on the correct label. Hence, to validate our method, we collect human judgments on the same ensemble of models and show that EigenBench's judgments align closely with those of human evaluators. We further demonstrate that EigenBench can recover model rankings on the GPQA benchmark without access to objective labels, supporting its viability as a framework for evaluating subjective values for which no ground truths exist.

In this paper, we propose a hybrid framework that heals corrupted finite semigroups, combining deterministic repair strategies with Machine Learning using a Random Forest Classifier. Corruption in these tables breaks associativity and invalidates the algebraic structure. Deterministic methods work for small cardinality n and low corruption but degrade rapidly. Our experiments, carried out on Mace4-generated data sets, demonstrate that our hybrid framework achieves higher healing rates than deterministic-only and ML-only baselines. At a corruption percentage of p=15%, our framework healed 95% of semigroups up to cardinality n=6 and 60% at n=10.