Temporal-aware consensus modeling for multi-aspect group satisfaction detection through fine-grained online review analysis

Fine-grained rehabilitation motion analysis: A knowledge-enhanced multi-branch network approach

Initiate-manage-sustain: A multimodal conversation analytic approach to understanding therapeutic engagement for children with autism spectrum disorder.

Fine-grained analysis of intraoperative behavior and its impact on patient outcomes remains a longstanding challenge. We present Frame-to-Outcome (F2O), an end-to-end system that translates tissue dissection videos into gesture sequences and uncovers patterns associated with postoperative outcomes. Leveraging transformer-based spatial and temporal modeling and frame-wise classification, F2O robustly detects consecutive short (˜2 s) gestures in the nerve-sparing step of robot-assisted radical prostatectomy (AUC: 0.80 frame-level; 0.81 video-level). F2O-derived features-gesture frequency, duration, and transitions-predicted postoperative outcomes with accuracy comparable to human annotations (0.79 vs. 0.75; overlapping 95% CI). Across 25 shared features, effect size directions were concordant with small differences (∆davg ≈ 0.07), and strong correlation (r = 0.96, p < 1 × 10-14). F2O also captured key patterns linked to erectile function recovery, including prolonged tissue peeling and reduced energy use. By enabling automatic interpretable assessment, F2O establishes a foundation for data-driven surgical feedback and prospective clinical decision support.

Evaluation of Image Quality Assessment (IQA) models has long been dominated by global correlation metrics, such as Pearson Linear Correlation Coefficient (PLCC) and Spearman Rank-Order Correlation Coefficient (SRCC). While widely adopted, these metrics reduce performance to a single scalar, failing to capture how ranking consistency varies across the local quality spectrum. For example, two IQA models may achieve identical SRCC values, yet one ranks high-quality images (related to high Mean Opinion Score, MOS) more reliably, while the other better discriminates image pairs with small quality/MOS differences (related to $|\Delta$MOS $|$). Such complementary behaviors are invisible under global metrics. Moreover, SRCC and PLCC are sensitive to test-sample quality distributions, yielding unstable comparisons across test sets. To address these limitations, we propose Granularity-Modulated Correlation (GMC), which provides a structured, fine-grained analysis of IQA performance. GMC includes: (1) a Granularity Modulator that applies Gaussian-weighted correlations conditioned on absolute MOS values and pairwise MOS differences ($|\Delta$ MOS$|$) to examine local performance variations, and (2) a Distribution Regulator that regularizes correlations to mitigate biases from non-uniform quality distributions. The resulting correlation surface maps correlation values as a joint function of MOS and $|\Delta$MOS$|$, providing a 3D representation of IQA performance. Experiments on standard benchmarks show that GMC reveals performance characteristics invisible to scalar metrics, offering a more informative and reliable paradigm for analyzing, comparing, and deploying IQA models. Codes are available at https://github.com/Dniaaa/GMC.

ABSTRACT How is attention distributed during reading? Using the MECO corpus, we report a first test of Parafoveal-on-Foveal (PoF) effects of orthographic overlap as a window on attention in natural reading settings. In contrast to previous studies, cross-linguistic comparisons across four alphabetic languages (English, Spanish, Finnish, Russian) revealed no uniform pattern: the well-established orthographic PoF effect from word N + 1 emerged only in Finnish and under specific constraints. More fine-grained analyses, testing orthographic overlap effects at different eccentricities showed significant effects for English and Finnish extending beyond word N + 1, especially around areas likely to comprise an inter-word space. Most of these effects were non-linear, reflecting both facilitation and disruption, depending on the extent of orthographic overlap. These results illustrate that, although PoF effects of orthographic overlap are decidedly more fragile in natural text reading, they still prevail with decidedly lower amounts of overlap and without artificial manipulations that may inadvertently bias attention.

Aortic arch calcification (AoAC) is commonly classified into four grades according to the percentage of calcification observed in clinical practice, and the interpretation is typically based on visual assessment by clinicians. However, this manual evaluation process is time-consuming and may fail to detect subtle calcifications, potentially leading to grading inaccuracies. This study presents a transformer-based model, termed Multi-Attention with Transformer Model (MATM), to improve the accuracy of AoAC grade classification. The proposed framework integrates multiple attention modules to enhance the representation of spatial features. In addition, the transformer mechanism incorporates positional information together with a hierarchical 16-block representation of the aortic arch, enabling fine-grained analysis of calcification distribution. The proposed method captures subtle calcification features and enables more accurate classification of AoAC grades. Experimental results demonstrate that the model can automatically estimate AoAC severity for both the traditional four-grade classification and the more detailed 16-grade classification, achieving an accuracy of up to 95.5%. The proposed method can reduce interpretation time and improve grading consistency for clinicians by minimizing variability caused by individual experience. Such AI-assisted assessment has the potential to standardize AoAC evaluation in future clinical practice.

SaBER-LLMs: Security-aware behavior embedding representation via large language models for fine-grained malware analysis

Fine-grained 3D phenotypic analysis of rice plays a vital role in rice breeding and yield estimation. However, a comprehensive rice data acquisition and segmentation pipeline is still lacking. While Neural Radiance Fields (NeRF) have shown impressive results in crop-level 3D reconstruction, their high sensitivity to data volume and camera viewpoints often leads to reconstruction failures for rice. In addition, the large-scale rice point clouds, coupled with heavy occlusion and visual similarity among grains, pose significant challenges for fine-grained trait extraction. To address the challenge of reconstructing rice point clouds under low-quality data conditions, we propose a novel method named Multi-Scale NeRF(MSNeRF). This method incorporates a structure-detail collaborative reconstruction mechanism and a dynamic initialization density scheduling strategy. Furthermore, we introduce a multimodal and multitask rice dataset (MMR) as a benchmark resource for future research. For rice point cloud segmentation, we develop Vision Rice Knowledge Graph Network(VRKGNet), which comprises an image segmentation module, a projection module, and a point cloud segmentation module enhanced with a Transformer to enlarge the receptive field. VRKGNet performs standalone point cloud segmentation and integrates image segmentation results from multiple viewpoints as prior knowledge to enhance semantic and instance-level segmentation. Extensive experiments demonstrate that MSNeRF achieves high-fidelity point cloud reconstruction with as few as 10 viewpoints. VRKGNet achieves superior rice plant segmentation with a semantic segmentation mIoU of 88.79% and an instance segmentation AP25 of 84.55%, outperforming mainstream algorithms.

Conflict tasks often yield smaller mean congruency effects when relevant (target) and irrelevant (distractor) information is mostly incongruent rather than mostly congruent. While this proportion congruency effect may reflect proactive control adaptation, only a few previous studies have provided convincing evidence for proactive control adaptation when ruling out contingency learning and reactive (item-specific) control adaptation. In this study, we present further evidence for proactive control adaptation (as reflected in proportion congruency effects and asymmetrical list-shifting effects in contingency-controlled diagnostic items) across three experiments (each N = 100 participants) using manual counting Stroop tasks (Experiment 1: number words as distractors; Experiment 2: Arabic numerals as distractors; Experiment 3: number words and Arabic numerals as inducer and diagnostic items or vice versa). To better understand the processes underlying proactive control adaptation, we conducted fine-grained distributional analyses (delta functions) and model-based analyses (diffusion model for conflict tasks). These analyses suggest that proactive control adaptation in manual counting Stroop tasks mainly reflects adjustments in the strength of distractor suppression rather than in the timing of distractor suppression, the strength of target amplification, or response caution. Additional distributional and diffusion model reanalyses of the data by Spinelli and Lupker (2023, Experiments 1-3) revealed a similar pattern in vocal color Stroop tasks. In conclusion, the present study provides new evidence for proactive control adaptation in manual counting Stroop tasks and indicates that proactive control adaptation mainly reflects adjustments in the strength of distractor suppression in both manual and vocal Stroop tasks. (PsycInfo Database Record (c) 2026 APA, all rights reserved).

IFSA-CE: Interpretable fine-grained sentiment analysis with concept embedding

Our understanding of sex differences in reward learning has been limited due to the predominant study of males. Here, we evaluated sex differences in flexible learning in two domains: the learning of stimulus-and action-based associations and their reversals. During action-based learning, rats selected between two identical visual stimuli presented on a touchscreen, where the spatial location predicted a higher probability of reward. For stimulus-based learning, rats chose between two distinct visual stimuli presented in pseudorandom spatial locations, one of which was associated with a higher probability of reward. Reversal phases involved switching reward contingency between the two actions or stimuli. We found that females did not differ in their discrimination or reversal learning accuracy compared to males, yet they collected fewer rewards and were more likely to omit trials than males in both domains. To gain a detailed understanding of differences across conditions, we modeled animals' trial-by-trial choices using reinforcement learning (RL) models and examined their steady-state behavior to capture transitions between distinct behavioral states. Although the estimated parameters of the best-fitting RL model revealed some sex differences, the model that incorporated transitions between different behavioral states provided a better overall fit to the data. This model also revealed that across all reversal phases, females exhibited a higher transition-specific lapse rate than males, indicating greater task disengagement once there was no need for further learning. Together, our fine-grained analysis of behavior adds to a growing literature on sex differences in flexible reward learning.

Deepfake detection has become an essential task for ensuring the authenticity and security of digital media. Although recent approaches have achieved notable progress, most existing detectors still exhibit limited generalization to unseen forgery techniques and remain vulnerable to common perturbations such as compression, noise, and adversarial attacks. To overcome these issues, we propose Weakly Supervised EfficientNet Augmented Face Forgery Detector (WAFF), a novel framework that integrates fine-grained per-frame analysis with adaptive video-level fusion. Specifically, WAFF integrates WSEffiNet, an EfficientNet-B3-based backbone enhanced with a Weakly Supervised Data Augmentation Network (WS-DAN). This design generates attention maps to emphasize subtle facial forgery artifacts while encouraging complementary local-global feature learning. At the video level, WAFF incorporates a multi-strategy fusion scheme that combines fake-frame counting, confidence averaging, and attention-guided voting to strike a balance between sensitivity and stability. Extensive experiments on FaceForensics++, Celeb-DF v2, DFD, DFDC, and FFIW-10K demonstrate that WAFF can achieve state-of-the-art performance under both high- and low-quality compression, while also enhancing cross-dataset generalization.

ABSTRACT This article analyzes the United States role in Bolivia’s 1967 operation to capture and execute Ernesto ‘Che’ Guevara. Existing literature is marked by three interpretations. One dismisses claims of CIA involvement as sensationalist and unfounded. Another, based on Bolivian military accounts, frames the operation as domestic and sovereign. A third perspective argues for US complicity, ranging from claims that the CIA ‘got away with murder’ to shadow involvement by American officials who ‘washed their hands’ of the affair. Drawing on newly available sources, we offer a fine-grained analysis of US intervention in the manhunt and execution.

Human rights violations reporting dataset

Fine-grained sentiment analysis of massive open online courses evaluation

• Integrating crop processing analysis (archaeobotany) and plant stable isotopic analysis to investigate agricultural practice and corresponding labor organization with finer resolution of prehistoric communities. • Millet cultivation at Bronze Age Erdaojingzi in Northeast China was likely organized at a household level, the products were shared among household members. • Direct evidence of the autonomy in decision-making and land tenure-ship within each household in the agricultural production of Bronze Age North China. • Morphological and isotopic evidence provide the circumstantial evidence of multi-seasonal harvesting regime of Chinese millet in ancient times. In this research, stable isotope analysis of plants was integrated with macrofossil studies to investigate social conditions underpinning millet cultivation and mundane food production at the Bronze Age village Erdaojingzi (3600–3500 cal. yr BP), Northeast China. Specifically, we sought to answer two questions: 1) Did different households (or other groupings) manage cultivation collectively in similar environments, or did house groups have unequal access to (or ownership of) different farmlands? 2) Were millet harvests organized according to house group distinctions, or were they equally distributed among households? How were crops harvested in the field within this social context? Our results seem to suggest that both cultivation and harvest were configured in the context of the household economy. Millet was cultivated in environments only accessible to certain house groups, while the products were shared among household members rather than pooled across the whole community. The correlation between morphologically attested grain sizes and isotopic results indicated circumstances that grains from the same family plot were harvested multiple times sequentially, rather than in a single seasonal harvesting regime. Extensive archaeobotanical sampling at Erdaojingzi has enabled fine-grained analysis of how millet cultivation and production were woven into social conditions operating both in the field and within the settlement.

Trusting selectively is a crucial ability in successfully navigating social environments: trusting a trustworthy partner maximizes mutual benefits; withholding trust from an untrustworthy partner minimizes chances of being exploited. Here, we ask how children's trust expectations and experienced trustworthiness inform their own trust decisions. Using a child-friendly repeated Trust Game, we employ a computational modeling approach to examine how children learn about the trustworthiness of others through experience and adjust their own trust behaviors. We tested N = 96 children ages 6-11 years who played 40 trials of the game as trustors with a trustworthy and an untrustworthy trustee. Regression analyses showed that overall, children shared more with trustworthy than untrustworthy partners and older children shared overall more than younger children. Computational models provided a more fine-grained analysis of the learning pattern: a reinforcement learning model that included parameters for individual variation in prior trust expectations and a parameter quantifying the behavioral adjustment over time provided the best fit for the data. Using this computational approach, we could ascertain that children's behavioral adjustments became less responsive over time, suggesting a shift toward more stable patterns of trust once initial impressions were formed. Together, these findings shed light on the development of children's trust interaction with novel partners and highlight the opportunity that computational modeling provides in capturing developmental differences in trust behaviors. SUMMARY: Children aged 6-11 used experience to distinguish trustworthy from untrustworthy partners in a repeated Trust Game, sharing more with the trustworthy than the untrustworthy partner. Older children shared more than younger children, although they were not significantly differentiating between the trustworthy than the untrustworthy partner. Reinforcement learning modeling suggested that children updated trust expectations from experience, with age-related differences in the efficiency of impression updating. These findings highlight both children's capacity in flexibly adjusting their economic trust behaviors and the value of computational modeling in studying economic trust development.

Abstract Building on Meyer, van Witteloostuijn, and Beugelsdijk’s (2017) editorial on best practices for conducting and reporting hypothesis-testing research in IB, we examine the extent to which their guidelines—and related recommendations by Hahn and Ang (2017)—have been adopted in leading IB journals. We analyze all null-hypothesis significance testing-based articles published in the Journal of International Business Studies and the Journal of World Business between 2012 and 2024, using fine-grained inferential trend analyses of methodological and reporting standards alongside state-of-the-art tests for p-hacking and publication bias. Our results indicate meaningful progress in several areas, including greater methodological rigor and transparency. However, adoption remains uneven and has plateaued for several practices. Persistent shortcomings include limited reporting of standard errors, confidence intervals, and effect sizes, incomplete disclosure of robustness analyses and outlier treatment, and the continued predominance of confirmed hypotheses. Moreover, we find no evidence that p-hacking or publication bias have declined over time. Drawing on these results, we outline actionable recommendations for advancing methodological and reporting standards in IB by (1) enhancing transparent reporting, (2) providing convincing evidence, and (3) reporting and probing null and negative results. Overall, our study offers a roadmap for strengthening research credibility in IB.

The policy recognition that peace and sustainability are intrinsically linked has led to increased scholarly attention on their nexus. In this systematic review, we focus on the indicators of positive peace and Sustainable Development Goals (SDGs) to provide a fine-grained analysis of their co-benefits, trade-offs, and mediating factors. The literature landscape shows an increasing conceptual integration of environmental, social, and economic issues, with the SDGs serving as the central organizing paradigm. Our content analysis also reveals that positive peace acts as the mediating factor, enabling the conditions or moderating the inadvertent consequences of sustainable development. In this age of multiple, interconnected, and fast-moving crises, advancing our understanding of the peace–sustainability nexus becomes more pressing than ever. This comprehensive assessment of the pathways between positive peace and sustainable development presents a synthesis of the dominant themes and identifies the knowledge gaps in the literature, thereby informing future research and policy directions.