Large language models (LLMs) are prone to hallucinations, i.e., statements unsupported by the input or training data, hindering reliable deployment. In parallel, numerous uncertainty estimation (UE) methods have been proposed to quantify model confidence and are often implicitly treated as proxies for model failure. However, the relationship between uncertainty and hallucinations remains insufficiently characterized. We present a systematic empirical study of the association between uncertainty estimators and hallucinations in LLMs. Rather than assuming this association, we evaluate directly when and to what extent it holds. We consider a diverse set of uncertainty estimators, including information-theoretic, sampling-based, and reflexive estimators, and examine their behavior across hallucination settings. Our experiments cover both intrinsic hallucinations (violations of input faithfulness) and extrinsic hallucinations (unsupported claims relative to training data), using four complementary benchmarks, including RAGTruth and HalluLens. We find that the association is highly variable and often weak, depending on the hallucination type and the LLM under evaluation. These results challenge the use of uncertainty as a direct signal of hallucination and clarify when it provides actionable information.

This paper presents the first benchmark of uncertainty disentanglement.

One paradigm of language model (LM) fine-tuning relies on creating large training datasets, under the assumption that high quantity and diversity will enable models to generalize to novel tasks after post-training. In practice, gathering large sets of data is inefficient, and training on them is prohibitively expensive; worse, there is no guarantee that the resulting model will handle complex scenarios or generalize better. Moreover, existing techniques rarely assess whether a training sample provides novel information or is redundant with the knowledge already acquired by the model, resulting in unnecessary costs. In this work, we explore a new test-time self-improvement method to create more effective and generalizable agentic LMs on-the-fly. The proposed algorithm can be summarized in three steps: (i) first it identifies the samples that model struggles with (self-awareness), (ii) then generates similar examples from detected uncertain samples (self-data augmentation), and (iii) uses these newly generated samples at test-time fine-tuning (self-improvement). We study two variants of this approach: Test-Time Self-Improvement (TT-SI), where the same model generates additional training examples from its own uncertain cases and then learns from them, and contrast this approach with Test-Time Distillation (TT-D), where a stronger model generates similar examples for uncertain cases, enabling student to adapt using distilled supervision. Empirical evaluations across different agent benchmarks demonstrate that TT-SI improves the performance with +5.48% absolute accuracy gain on average across all benchmarks and surpasses other standard learning methods, yet using 68x less training samples. Our findings highlight the promise of TT-SI, demonstrating the potential of self-improvement algorithms at test-time as a new paradigm for building more capable agents toward self-evolution.

Lower limb exoskeletons show promise to assist human movement, but their utility is limited by controllers designed for discrete, predefined actions in controlled environments, restricting their real-world applicability. We present an uncertainty-aware control framework that enables ankle exoskeletons to operate safely across diverse scenarios by automatically disengaging when encountering unfamiliar movements. Our approach uses an uncertainty estimator to classify movements as similar (in-distribution) or different (out-of-distribution) relative to actions in the training set. We evaluated three architectures (model ensembles, autoencoders, and generative adversarial networks) on an offline dataset and tested the strongest performing architecture (ensemble of gait phase estimators) online. The online test demonstrated the ability of our uncertainty estimator to turn assistance on and off as the user transitioned between in-distribution and out-of-distribution tasks (F1: 89.2). This new framework provides a path for exoskeletons to safely and autonomously support human movement in unstructured, everyday environments.

Thank you for running our software. R3: "This is so far the only code I was able to run among the ones I

We continue the derivation from the main here in more detail. It has been shown that (c.f. Where we utilize Woodbury's Identity [ ''' model: network trained with anchoring anchors: set of randomly chosen anchors (ideally from train dist.) B.2 ImageNet-C corruptions for OOD and Calibration Table 1 lists the set of corruptions used to construct the ImageNet-C benchmark. For each case, we show the negative log-likelihood for the test data obtained using each of the methods.

While uncertainty estimation for graphs recently gained traction, most methods rely on homophily and deteriorate in heterophilic settings. We address this by analyzing message passing neural networks from an information-theoretic perspective and developing a suitable analog to data processing inequality to quantify information throughout the model's layers. In contrast to non-graph domains, information about the node-level prediction target can increase with model depth if a node's features are semantically different from its neighbors. Therefore, on heterophilic graphs, the latent embeddings of an MPNN each provide different information about the data distribution - different from homophilic settings. This reveals that considering all node representations simultaneously is a key design principle for epistemic uncertainty estimation on graphs beyond homophily. We empirically confirm this with a simple post-hoc density estimator on the joint node embedding space that provides state-of-the-art uncertainty on heterophilic graphs. At the same time, it matches prior work on homophilic graphs without explicitly exploiting homophily through post-processing.

Accurately estimating image quality and model robustness improvement are critical challenges in unconstrained face recognition, which can be addressed through uncertainty estimation via probabilistic face embeddings. Previous research mainly focused on uncertainty estimation in face verification, leaving the open-set face recognition task underexplored. In open-set face recognition, one seeks to classify an image, which could also be unknown. Here, the low variance of probabilistic embedding does not imply a low error probability: an image embedding could be close to several classes in a gallery, thus yielding high uncertainty. We propose a method aware of two sources of ambiguity in the open-set recognition system: (1) the gallery uncertainty caused by overlapping classes and (2) the uncertainty of the face embeddings. To detect both types, we use a Bayesian probabilistic model of embedding distribution, which provides a principled uncertainty estimate. Challenging open-set face recognition datasets, such as IJB-C, serve as a testbed for our method. We also propose a new open-set recognition protocol for whale and dolphin identification. The proposed approach better identifies recognition errors than uncertainty estimation methods based solely on image quality.

Active learning (AL) is designed to construct a high-quality labeled dataset by iteratively selecting the most informative samples. Such sampling heavily relies on data representation, while recently pre-training is popular for robust feature learning. However, as pre-training utilizes low-level pretext tasks that lack annotation, directly using pre-trained representation in AL is inadequate for determining the sampling score. To address this problem, we propose a downstream-pretext domain knowledge traceback (DOKT) method that traces the data interactions of downstream knowledge and pre-training guidance for selecting diverse and instructive samples near the decision boundary. DOKT consists of a traceback diversity indicator and a domain-based uncertainty estimator. The diversity indicator constructs two feature spaces based on the pre-training pretext model and the downstream knowledge from annotation, by which it locates the neighbors of unlabeled data from the downstream space in the pretext space to explore the interaction of samples. With this mechanism, DOKT unifies the data relations of low-level and high-level representations to estimate traceback diversity. Next, in the uncertainty estimator, domain mixing is designed to enforce perceptual perturbing to unlabeled samples with similar visual patches in the pretext space. Then the divergence of perturbed samples is measured to estimate the domain uncertainty. As a result, DOKT selects the most diverse and important samples based on these two modules. The experiments conducted on ten datasets show that our model outperforms other state-of-the-art methods and generalizes well to various application scenarios such as semantic segmentation and image captioning.