The Domain Name System (DNS) is responsible for translating human-readable names into computer-friendly identifiers and employs a reverse tree architecture to facilitate this translation. The forward DNS (fDNS) translates human-readable domain names into computer values, while the reverse DNS (rDNS) usually maps an IP address to a hostname. The DNS has evolved into a more intricate system, driven by economic incentives, with data stewardship being delegated to numerous autonomous stakeholders. Due to its role in Internet infrastructure, studying DNS is fundamental for both practical and security reasons. In this work, we characterize commonly used DNS datasets and the challenges linked to their usage. We thoroughly survey DNS focused research studies published over the course of more than a decade, within more than 15 proceedings, and observe a prominent reliance (62%) on short-term datasets. Our investigation into the challenges that limit the adoption of long-term datasets reveals that some of these challenges are associated with the distributed design of the DNS, while others are associated with the evolution of the DNS deployment. That is, we found that, the diversity of stakeholders has the potential to impede data collection and access by various stakeholders, as well as the sharing of data between operators and researchers. Specifically, the distributed design of the DNS leads to challenges regarding namespace coverage, privacy, confidentiality, format, size, and time granularity. Furthermore, certain DNS data is considered commercially sensitive, which further increases the penumbra over the global DNS exchanges, leading to a lack of publicly available DNS data. In earlier works, we observe a distinction in research topics between studies using actively and passively collected DNS data, followed by a recent convergence in topics investigated across both data collection approaches. Lastly, we noted that rDNS data is studied much less than fDNS, i.e., only 2.7% of the studied papers.
Missing data is a prevalent issue that can significantly impair model performance and explainability. This paper briefly summarizes the development of the field of missing data with respect to Explainable Artificial Intelligence and experimentally investigates the effects of various imputation methods on SHAP (SHapley Additive exPlanations), a popular technique for explaining the output of complex machine learning models. Next, we compare different imputation strategies and assess their impact on feature importance and interaction as determined by Shapley values. Moreover, we also theoretically analyze the effects of missing values on Shapley values. Importantly, our findings reveal that the choice of imputation method can introduce biases that could lead to changes in the Shapley values, thereby affecting the explainability of the model. Moreover, we also show that a lower test prediction MSE (Mean Square Error) does not necessarily imply a lower MSE in Shapley values and vice versa. Also, while XGBoost (eXtreme Gradient Boosting) is a method that could handle missing data directly, using XGBoost directly on missing data can seriously affect explainability compared to imputing the data before training XGBoost. This study provides a comprehensive evaluation of imputation methods in the context of model explanations, offering practical guidance for selecting appropriate techniques based on dataset characteristics and analysis objectives. The results underscore the importance of considering imputation effects to ensure robust and reliable insights from machine learning models.
Hallucinations in video-capable vision-language models (Video-VLMs) remain frequent and high-confidence, while existing uncertainty metrics often fail to align with correctness. We introduce VideoHEDGE, a modular framework for hallucination detection in video question answering that extends entropy-based reliability estimation from images to temporally structured inputs. Given a video-question pair, VideoHEDGE draws a baseline answer and multiple high-temperature generations from both clean clips and photometrically and spatiotemporally perturbed variants, then clusters the resulting textual outputs into semantic hypotheses using either Natural Language Inference (NLI)-based or embedding-based methods. Cluster-level probability masses yield three reliability scores: Semantic Entropy (SE), RadFlag, and Vision-Amplified Semantic Entropy (VASE). We evaluate VideoHEDGE on the SoccerChat benchmark using an LLM-as-a-judge to obtain binary hallucination labels. Across three 7B Video-VLMs (Qwen2-VL, Qwen2.5-VL, and a SoccerChat-finetuned model), VASE consistently achieves the highest ROC-AUC, especially at larger distortion budgets, while SE and RadFlag often operate near chance. We further show that embedding-based clustering matches NLI-based clustering in detection performance at substantially lower computational cost, and that domain fine-tuning reduces hallucination frequency but yields only modest improvements in calibration. The hedge-bench PyPI library enables reproducible and extensible benchmarking, with full code and experimental resources available at https://github.com/Simula/HEDGE#videohedge .
Large language models (LLMs) are increasingly used for decision support in healthcare, but clinical evidence is often incomplete or evolving. When the available information is insufficient to support a reliable answer, models should request clarification or abstain rather than provide unsupported responses. Existing medical benchmarks, however, typically assume that complete information is available upfront. We introduce Safe-Psych, a sequential benchmark for evaluating how LLMs handle evolving diagnostic uncertainty in clinical psychiatry. Safe-Psych contains over 1,000 real-world psychiatric clinical notes segmented to simulate incremental evidence disclosure, with psychiatrist-derived action labels at each stage: DIAGNOSE, CLARIFY, or ABSTAIN. We evaluate multiple state-of-the-art LLMs in full-information and sequential settings. Our findings show that capability does not ensure calibration: even strong models struggle under incomplete clinical information, with under-abstention exceeding 60
Large Language Models (LLMs) rely on various decoding strategies to generate text, and these choices can significantly affect output quality. In healthcare, where accuracy is critical, the impact of decoding strategies remains underexplored. We investigate this effect in five open-ended medical tasks, including translation, summarization, question answering, dialogue, and image captioning, evaluating 11 decoding strategies with medically specialized and general-purpose LLMs of different sizes. Our results show that deterministic strategies generally outperform stochastic ones: beam search achieves the highest scores, while η and top-k sampling perform worst. Slower decoding methods tend to yield better quality. Larger models achieve higher scores overall but have longer inference times and are no more robust to decoding. Surprisingly, while medical LLMs outperform general ones in two of the five tasks, statistical analysis shows no overall performance advantage and reveals greater sensitivity to decoding choice. We further compare multiple evaluation metrics and find that correlations vary by task, with MAUVE showing weak agreement with BERTScore and ROUGE, as well as greater sensitivity to the decoding strategy. These results highlight the need for careful selection of decoding methods in medical applications, as their influence can sometimes exceed that of model choice.
