Search Results (91)

The evaluation of large language models (LLMs) increasingly relies on other LLMs acting as automated judges. While this approach offers scalability and efficiency, it raises serious concerns regarding evaluator reliability, positional bias, and ranking stability. This paper presents a scalable framework for diagnosing positional bias and instability in LLM-based evaluation by using controlled pairwise comparisons judged by multiple independent language models. The system supports mirrored comparisons with reversed response order, prompt injection, and surface-level perturbations (e.g., paraphrasing, lexical noise), enabling fine-grained analysis of evaluator consistency and verdict robustness. Over 3600 pairwise comparisons were conducted across five instruction-tuned open-weight models using ten open-ended prompts. The top-performing model (gemma:7b-instruct) achieved a 66.5% win rate. Evaluator agreement was uniformly high, with 100% consistency across judges, yet 48.4% of verdicts reversed under mirrored response order, indicating strong positional bias. Kendall’s Tau analysis further showed that local model rankings varied substantially across prompts, suggesting that semantic context influences evaluator judgment. All evaluation traces were stored in a graph database (Neo4j), enabling structured querying and longitudinal analysis. The proposed framework provides not only a diagnostic lens for benchmarking models but also a blueprint for fairer and more interpretable LLM-based evaluation. These findings underscore the need for structure-aware, perturbation-resilient evaluation pipelines when benchmarking LLMs. The proposed framework offers a reproducible path for diagnosing evaluator bias and ranking instability in open-ended language tasks. Future work will apply this methodology to educational assessment tasks, using rubric-based scoring and graph-based traceability to evaluate student responses in technical domains. Full article

This article examines LGBTQI+ asylum claims in the U.S. Circuit Court of Appeals. The data are part of a larger study that has identified 520 LGBTQI+ claims in the U.S. Circuit of Appeals from 1994 to 2023. It focuses on examples from the 115 cases that were granted a review and analyzes the logic that U.S. Circuit Court justices use when deciding to grant a review of a petition that was denied by a lower court, such as the Board of Immigration Appeals (BIA) and immigration courts. This article argues that the U.S. Circuit of Appeals contests lower court rulings from BIA and immigration court judges based on assumptions about credibility, discretion, persecution, and criminalization for LGBTQI+ asylum seekers. By granting reviews, the Circuit Courts provide an opening for the acceptance of queer asylum claims. Full article

The interaction between antiresorptive medication and dental implant procedures remains a subject of concern complicating the decision-making process for clinicians. The aim of the study is to conduct a literature review on the relationship between dental implant placement and the incidence of osteonecrosis of the jaw (ONJ) in patients receiving antiresorptive drugs. The systematic review relied on the PRISMA statement using the PICO tool. The literature search was performed using PubMed, EBSCOhost and Scopus for RCTs, controlled clinical trials and cohort studies. The choice of reference studies was made in a blind process with a 100% agreement rate. For all included studies, quality assessment was performed. The research led to the selection of 608 results. Only five studies were included in the review. Three of the included studies were judged as having a low risk of bias. Dental implants may not be linked to a higher risk of osteonecrosis of the jaw in patients taking low-dose bone-modifying agents. The long-term survival of implants in osteoporotic patients taking oral antiresorptive medication was similar to that in a healthy population and significantly higher than in untreated controls. Full article

Cognitive bias is widespread, hidden, and difficult to deal with. It impacts each and every aspect of the justice and legal systems, from the initial engagement of police officers attending the crime scene, through the forensic examination, and all the way to the final outcome of the jurors’ verdict and the judges’ sentencing. It impacts not only the subjective elements in the justice and legal systems but also the more objective scientific elements, such as forensic fingerprinting and DNA. The impact of bias on each of these elements has mainly been researched and considered in silo, neglecting the biasing interactions and how bias cascades and snowballs throughout the justice and legal systems. These should happen rarely, as the Swiss cheese model shows that such errors in the final outcome rarely occur because they require that the shortcomings in each element be coordinated and aligned with the other elements. However, in the justice and legal systems, the different elements are not independent; they are coordinated and mutually support and bias each other, creating and enabling hidden bias cascade and bias snowball effects. Hence, minimizing bias requires not only taking measures to reduce bias in each of the elements but also a wider perspective that addresses bias cascade and bias snowball effects. Full article

Large language models (LLMs) are increasingly being utilized to develop tools and services in various domains, including education. However, due to the nature of the training data, these models are susceptible to inherent social or cognitive biases, which can influence their outputs. Furthermore, their handling of critical topics, such as privacy and sensitive questions, is essential for responsible deployment. This study proposes a framework for the automatic detection of biases and violations of responsible use using a synthetic question-based dataset mimicking student–chatbot interactions. We employ the LLM-as-a-judge method to evaluate multiple LLMs for biased responses. Our findings show that some models exhibit more bias than others, highlighting the need for careful consideration when selecting models for deployment in educational and other high-stakes applications. These results emphasize the importance of addressing bias in LLMs and implementing robust mechanisms to uphold responsible AI use in real-world services. Full article

Background: Given the relationship between reduced pulmonary and respiratory muscle function in neck pain, incorporating breathing exercises into neck pain management may be beneficial. Purpose: The purpose of this study was to investigate the benefits of breathing exercises for treating neck pain. Methods: We searched PubMed (MEDLINE), PEDro, CINAHL, Scopus, and EMBASE databases, up to the 28 of February 2024. Randomized controlled trials evaluating the impact of breathing exercises on reducing pain and disability in both persistent and recent neck pain were selected. A meta-analysis was conducted for each outcome of interest; however, if quantitative methods were not possible, a qualitative synthesis approach was used. The risk of bias was assessed using the Cochrane RoB 2.0 Tool (version 22 August 2019). We used the GRADE approach to judge the certainty of the evidence. Results: Five studies were included. Meta-analysis showed a statistically significant reduction in pain (standardized mean difference (SMD), −10.16; 95% CI: −14.82, −5.50) and disability (SMD, −0.80; 95% CI: −1.49, −0.11), in favor of breathing exercises. Qualitative synthesis for pulmonary functional parameters resulted in a statistically significant improvement for FVC, MIP, MEP, and MVV, in favor of breathing exercises. Conclusions: Breathing exercises showed significant short-term effects in reducing pain and disability for persistent neck pain. They also provided benefits for functional respiratory parameters. However, the evidence certainty is low. Full article

Background/Objectives: This systematic review aimed to investigate the efficacy of methylphenidate medication in the treatment of cognitive problems, such as attention, following pediatric traumatic brain injury. Previous reviews have focused on a broader population of acquired brain injury in pediatrics. Methods: Six databases were systematically searched, and eleven relevant reports were included, of which five were randomised controlled trials (RCTs) and six were prospective cohort designs with no control arm. The risk of bias was assessed for each of the studies using appropriate tools. Results: Eleven studies were included in this study for data extraction consisting of 376 participants. Our primary outcome of the efficacy of methylphenidate in improving attention was assessed in the included studies using a variety of tools. A meta-analysis was only possible for studies using the continuous performance test data, which showed an overall insignificant reduction of 36.07 (95% CI [−96.94, 24.80], p = 0.25). Other outcomes, such as the Conners’ rating scale and the behaviour rating interview of executive function, also did not show an overall difference after methylphenidate treatment. However, the risk of bias across all studies was judged as moderate to high. Conclusions: We conclude that there is currently no evidence to support the use of methylphenidate to improve cognitive outcomes in pediatric traumatic brain injury patients. Significantly larger high-quality studies are needed to determine an effect on executive functioning outcomes after methylphenidate treatment in pediatric traumatic brain injury. Full article

Background: Nutritional supplements are widely used by swimmers, but the effectiveness of various supplements and the identification of the most effective intervention require further investigation. Purpose: This paper evaluated and compared the effectiveness of various nutrition-based interventions on swimming performance through both direct and indirect comparisons. Methods: PubMed, Embase, Web of Science, Cochrane Library, and SPORTDiscus databases were thoroughly searched up to 4 April 2024. The risk of bias was judged using the Cochrane risk of bias tool. A random-effect model was adopted to compute standardized mean differences (SMD) and 95% confidence intervals (CI). Results: L-arginine (Arg) demonstrated superior performance to the placebo (SMD = −1.66, 95% CI [−2.92, −0.44]), emerging as the most effective intervention for reducing 100 swimming time (SUCRA = 89.5%). Beta-alanine (BA) was the best intervention for improving blood lactate (SUCRA = 80%). Creatine combined with sodium bicarbonate (Creatine_NaHCO₃) significantly increased blood pH compared to the placebo (SMD = 3.79, 95% CI [1.85, 5.80]), with a SUCRA score of 99.9%, suggesting it is the most effective intervention for this parameter. No prominent differences were noted among the interventions in 50 m time, 200 m time, heart rate, and body mass. Conclusions: Dietary supplements might provide benefits for improving swimming performance. Arg emerged as the most efficacious modality for reducing 100 m time. BA proved to be the preeminent strategy for decreasing blood lactate. Creatine_NaHCO₃ was distinguished as the optimal approach for improving blood pH. Full article

We identify and present Brazil’s most common behavioral and heuristic biases in judicial decision-making. Through bibliographic and specific cases, we notice the occurrence of the representativeness heuristic, availability heuristic, anchoring heuristic (anchoring effect), confirmation bias, and affect heuristic bias in Brazilian judicial decisions. We also present the current state of Brazilian legislation and its amendments that aim at impartiality in the production, the assessment of evidence, and the judge’s conviction. Finally, we present the suggestions and initiatives that aim to mitigate biases and heuristics in judicial decision-making in Brazil, especially with awareness techniques, the replacement of judges by algorithms, and the review of judicial decisions by collegiate bodies. Full article

Light field images can record multiple information about the light rays in a scene and provide multiple views from a single image, offering a new data source for 3D reconstruction. However, ensuring the quality of light field images themselves is challenging, and distorted image inputs may lead to poor reconstruction results. Accurate light field image quality assessment can pre-judge the quality of light field images used as input for 3D reconstruction, providing a reference for the reconstruction results before the reconstruction work, significantly improving the efficiency of 3D reconstruction based on light field images. In this paper, we propose an Adaptive Vision Transformer-based light-field image-quality assessment model (AViT-LFIQA). The model adopts a multi-view sub-aperture image sequence input method, greatly reducing the number of input images while retaining as much information as possible from the original light field image, alleviating the training pressure on the neural network. Furthermore, we design an adaptive learnable attention layer based on ViT, which addresses the lack of inductive bias in ViT by using adaptive diagonal masking and a learnable temperature coefficient strategy, making the model more suitable for training on small datasets of light field images. Experimental results demonstrate that the proposed model is effective for various types of distortions and shows superior performance in light-field image-quality assessment. Full article

Precisely estimating the position, diameter at breast height (DBH), and height of trees is essential in forest resource inventory. Augmented reality (AR)-based devices help overcome the issue of inconsistent global point cloud data under thick forest canopies with insufficient Global Navigation Satellite System (GNSS) coverage. Although monocular simultaneous localization and mapping (SLAM) is one of the current mainstream systems, there is still no monocular SLAM solution for forest resource inventories, particularly for the precise measurement of inclined trees. We developed a forest plot survey system based on monocular SLAM that utilizes array cameras and Inertial Measurement Unit (IMU) sensors provided by smartphones, combined with augmented reality technology, to achieve a real-time estimation of the position, DBH, and height of trees within forest plots. Our results from the tested plots showed that the tree position estimation is unbiased, with an RMSE of 0.12 m and 0.11 m in the x-axis and y-axis directions, respectively; the DBH estimation bias is −0.17 cm (−0.65%), with an RMSE of 0.83 cm (3.59%), while the height estimation bias is −0.1 m (−0.95%), with an RMSE of 0.99 m (5.38%). This study will be useful in designing an algorithm to estimate the DBH and position of inclined trees using point clouds constrained by sectional planes at the breast height of the trunk, developing an algorithm to estimate the height of inclined trees utilizing the relationship between rays and plane positions, and providing observers with visual measurement results using augmented reality technology, allowing them to judge the accuracy of the estimates intuitively. Clearly, this system has significant potential applications in forest resource management and ecological research. Full article

We assessed the aesthetic experience of patients with behavioral variant frontotemporal dementia (bvFTD) to understand their ability to experience feelings of the sublime and to be moved when viewing paintings. We exposed patients with bvFTD and control participants to concrete and abstract paintings and asked them how moved they were by these paintings and whether the latter were beautiful or ugly. Patients with bvFTD declared being less moved than control participants by both abstract and concrete paintings. No significant differences were observed between abstract and concrete paintings in both patients with bvFTD and control participants. Patients with bvFTD provided fewer “beautiful” and more “ugly” responses than controls for both abstract and concrete paintings. No significant differences in terms of “beautiful” and “ugly” responses were observed between abstract and concrete paintings in both patients with bvFTD and control participants. These findings suggest disturbances in the basic affective experience of patients with bvFTD when they are exposed to paintings, as well as a bias in their ability to judge the aesthetic quality of paintings. Full article

The activity of dental caries, combined with its multifactorial etiology, alters salivary molecule composition. The present systematic review was developed to answer the following question: “Are salivary biomarkers reliable for diagnosis of dental caries?”. Following the “Preferred Reporting Item for Systematic Reviews and Meta-analysis” (PRISMA) guidelines, the review was conducted using multiple database research (Medline, Web of Science, and Scopus). Studies performed on healthy subjects with and without dental caries and providing detailed information concerning the clinical diagnosis of caries (Decayed, Missing, Filled Teeth-DMFT and International Caries Detection and Assessment System-ICDAS criteria) were included. The quality assessment was performed following a modified version of the Joanna Briggs Institute Prevalence Critical Appraisal Checklist. The protocol was registered in the International Prospective Register of Systematic Reviews (PROSPERO, ID: CRD42022304505). Sixteen papers were included in the review. All studies reported statistically significant differences in the concentration of salivary molecules between subjects with and without caries (p < 0.05). Proteins were the most investigated molecules, in particular alpha-amylase and mucins. Some studies present a risk of bias, such as identifying confounding factors and clearly defining the source population. Nevertheless, the 16 papers were judged to be of moderate to high quality. There is evidence that some salivary compounds studied in this review could play an important diagnostic role for dental caries, such as salivary mucins, glycoproteins (sCD14), interleukins (IL-2RA, 4,-13), urease, carbonic anhydrase VI, and urea. Full article

Background: Surface electromyography (sEMG) can provide an objective and quantitative image of the functional state of neuromuscular balance in the stomatognathic system. The objective of this systematic review is to examine current scientific evidence regarding the effects of orthodontic treatment on muscle electromyographic (EMG) activity in children. Methods: The search strategy included the PubMed, PubMed Central, Web of Science, Scopus, and Embase databases. The inclusion criteria were studies assessing EMG muscle activity in children undergoing orthodontic treatment compared with untreated children. The Cochrane risk-of-bias tool (RoB2) and the Newcastle–Ottawa Scale (NOS) were used to evaluate the quality of the studies. The quality of evidence assessment was performed using GRADE analysis. The PRISMA diagram visually represented the search strategy, as well as screening and inclusion process. Results: The search strategy identified 540 potential articles. Fourteen papers met the inclusion criteria. Six studies were judged at a low risk of bias. The certainty of evidence was rated as moderate to low, according to the GRADE criteria. Studies showed alterations in EMG muscle activity in children undergoing orthodontic treatment. Conclusions: Orthodontic treatment appears to affect muscle activity in children undergoing orthodontic treatment. However, the quality of evidence is low and, therefore, it is not possible to definitively state this effect. Further long-term studies are needed to confirm the findings of this review. Study protocol number in PROSPERO database: CRD42023491005. Full article

Gestational diabetes mellitus (GDM) affects approximately 5–20% of pregnant women and is associated with adverse pregnancy outcomes. This review aimed to assess whether the oral microbiota of infants and their mothers with GDM had a different composition from that found in unaffected women and offspring. PubMed, Embase, Scopus, and Google Scholar were searched in December 2023 after protocol registration in the International Prospective Register of Systematic Reviews (CRD42023406505). Risk of bias was assessed using the Joanna Briggs Institute Critical Appraisal tools. Overall, 1113 articles were identified; after evaluating the full texts, 12 papers were included in the qualitative analysis. In six studies of the eight included, significant differences in microbiota between M-GDM and M-nGDM were found. In four studies, a depletion of Firmicutes and an enrichment of Proteobacteria was found in the microbiota of infants. Since all included studies were judged to have high risk of bias, a quantitative synthesis of the results was not carried out. In conclusion, although the oral microbiota of infants from mothers with GDM could be different from that of infants from mothers without GDM, there is insufficient evidence to clarify this aspect so far. Full article