A One-Shot Shell-HDR Conversion for Extremely Low-Lightness Object Awareness

Frontiers in artificial intelligence and applications · 2026-02-04

This work proposes a novel single shot enhancing low-light HDR approach, namely shell-HDR, to address limitations in night-scenario object detection. Unlike traditional multi-exposure HDR prone to motion artifacts, shell-HDR synthetically blends a single raw low-light image with its Histogram Equalization (HE) and Adaptive HE (AHE) processed versions. This method eliminates ghosting while preserving object positional consistency, creating an optimally enhanced image for labeling. The processed dataset, comprising 5,148 raw, HE, AHE, and shell-HDR images, then trains and validates comprehensive YOLO-series models (v8/v9/v10, s/m/l/x). Experimental results benchmark preprocessing/inference times, NMS latency, and mAP scores, demonstrating shell-HDR’s efficacy for robust visual intelligence in challenging lighting conditions compared to conventional enhancement techniques.

Human–Algorithmic Bias: Source, Evolution, and Impact

Management Science · 2025-09-09 · 4 citations

Prior work on human-algorithmic bias has seen difficulty in empirically identifying the underlying mechanisms of bias because in a typical “one-time” decision-making scenario, different mechanisms generate the same patterns of observable decisions. In this study, leveraging a unique repeat decision-making setting in a high-stakes microlending context, we aim to uncover the underlying source, evolution dynamics, and associated impacts of bias. We first develop a structural econometric model of the decision dynamics to understand the source and evolution of bias in human evaluators in microloan granting. We find that both preference-based and belief-based biases exist in human decisions and are in favor of female applicants. Our counterfactual simulations show that the elimination of either of the two biases improves the fairness in financial resource allocation as well as the platform profits. The profit improvement mainly stems from the increased approval probability for male borrowers, especially those who would eventually pay back loans. Furthermore, to examine how human biases evolve when being inherited by machine learning (ML) algorithms, we train state-of-the-art ML algorithms for default risk prediction on both real-world data sets with human biases encoded within and counterfactual data sets with human biases partially or fully removed. We find that even fairness-unaware ML algorithms can reduce bias in human decisions. Interestingly, although removing both types of human bias from the training data can further improve ML fairness, the fairness-enhancing effects vary significantly between new and repeat applicants. Based on our findings, we discuss how to reduce decision bias most effectively in a human-ML pipeline. This paper was accepted by D. J. Wu, Special Issue on the Human-Algorithm Connection. Supplemental Material: The online appendix and data files are available at https://doi.org/10.1287/mnsc.2022.03862 .

Turning Triangulation into Opportunity: A Dual-Adversary Algorithm for Fair and Profitable Credit Risk Modeling

SSRN Electronic Journal · 2025-01-01

PRVQL: Progressive Knowledge-Guided Refinement for Robust Egocentric Visual Query Localization

2025-10-19

Egocentric visual query localization (EgoVQL) focuses on localizing the target of interest in space and time from first-person videos, given a visual query. Despite recent progressive, existing methods often struggle to handle severe object appearance changes and cluttering background in the video due to lacking sufficient target cues, leading to degradation. Addressing this, we introduce PRVQL, a novel Progressive knowledge-guided Refinement framework for EgoVQL. The core is to continuously exploit target-relevant knowledge directly from videos and utilize it as guidance to refine both query and video features for improving target localization. Our PRVQL contains multiple processing stages. The target knowledge from one stage, comprising appearance and spatial knowledge extracted via two specially designed knowledge learning modules, are utilized as guidance to refine the query and videos features for the next stage, which are used to generate more accurate knowledge for further feature refinement. With such a progressive process, target knowledge in PRVQL can be gradually improved, which, in turn, leads to better refined query and video features for localization in the final stage. Compared to previous methods, our PRVQL, besides the given object cues, enjoys additional crucial target information from a video as guidance to refine features, and hence enhances EgoVQL in complicated scenes. In our experiments on challenging Ego4D, PRVQL achieves state-of-the-art result and largely surpasses other methods, showing its efficacy. Our code, model and results will be released at https://github.com/fb-reps/PRVQL.

Operational Decision Making Around Movie Piracy and Theatrical Release: A Structural Model of Movie Piracy vs. Legal (in-Theater) Consumption

Manufacturing & Service Operations Management · 2025-11-17

Problem definition: We examine moviegoers’ choices between consuming content via legal theatrical channels and illegal piracy channels. We focus on how two important factors affect this choice: the picture quality of piracy sources (a function of studio security investments) and the costs associated with legal channels, including consumer transportation costs (a function of screening volumes) and ticket prices. Methodology/results: We formulate a structural model and conduct counterfactual simulations. Our findings indicate that consumers’ choice between legal and illegal channels is significantly influenced by the quality of pirated sources and the costs of legal consumption. The emergence of high-quality piracy sources in the first week of a movie’s theatrical release leads to a 7.9% reduction in theatrical revenue during the first eight weeks of release, as compared with a scenario where only low-quality pirated sources are available. We also find that high-quality piracy sources pose a greater threat to the sales of smaller movies than of blockbusters. Managerial implications: Our work provides a rare insight into the operational decisions faced by movie studios around the theatrical delivery of movies. Our counterfactual simulations explore the potential for movie studios to manipulate the supply, cost, and quality associated with legal content to mitigate piracy and increase profit. Our results indicate that the potential for cost reductions is limited, as changes in ticket prices or screen volume have minimal impact on legal consumption. However, a moderate improvement in the value or quality of the theatrical experience (e.g., via one-time investment by theaters in upgrading equipment or technology) can effectively offset the impact of high-quality pirated content. Supplemental Material: The online appendix is available at https://doi.org/10.1287/msom.2023.0407 .

Dyn-D$^2$P: Dynamic Differentially Private Decentralized Learning with Provable Utility Guarantee

ArXiv.org · 2025-05-10

Most existing decentralized learning methods with differential privacy (DP) guarantee rely on constant gradient clipping bounds and fixed-level DP Gaussian noises for each node throughout the training process, leading to a significant accuracy degradation compared to non-private counterparts. In this paper, we propose a new Dynamic Differentially Private Decentralized learning approach (termed Dyn-D$^2$P) tailored for general time-varying directed networks. Leveraging the Gaussian DP (GDP) framework for privacy accounting, Dyn-D$^2$P dynamically adjusts gradient clipping bounds and noise levels based on gradient convergence. This proposed dynamic noise strategy enables us to enhance model accuracy while preserving the total privacy budget. Extensive experiments on benchmark datasets demonstrate the superiority of Dyn-D$^2$P over its counterparts employing fixed-level noises, especially under strong privacy guarantees. Furthermore, we provide a provable utility bound for Dyn-D$^2$P that establishes an explicit dependency on network-related parameters, with a scaling factor of $1/\sqrt{n}$ in terms of the number of nodes $n$ up to a bias error term induced by gradient clipping. To our knowledge, this is the first model utility analysis for differentially private decentralized non-convex optimization with dynamic gradient clipping bounds and noise levels.

Pandora Box or Golden Fleece: Economic Analysis of Generative AI Adoption on Creation Platforms

Proceedings of the ... Annual Hawaii International Conference on System Sciences/Proceedings of the Annual Hawaii International Conference on System Sciences · 2025-01-01 · 2 citations

In this paper, we employ a game-theoretic approach to analyze the impact of adopting generative AI tools on various market outcomes, including creators’ creativity level and pricing strategies, the platform’s regulatory strength, and consumer welfare. Our results show that although generative AI tools can facilitate content creation by leading to a higher quality, creators’ creativity levels and prices may experience a decline compared to the regime without such tools. We also find that even though generative AI has the potential to improve content quality, it is not always true that the utilization of such tools can lead to higher consumer welfare. Moreover, we find that the platform opts for a relatively low regulatory strength when the AI intelligence is relatively small. Interestingly, as generative AI acquires a more extensive knowledge base, the platform and creators may be less willing to adopt it.

Personalization, Consumer Search, and Algorithmic Pricing

Marketing Science · 2025-05-07 · 6 citations

This paper shows that personalized product rankings, although improving search relevance, can unintentionally enable AI pricing algorithms to raise prices and reduce consumer welfare.

A study on emotion recognition of autistic children integrating emotional intelligence and behavioral analysis

2025-04-11

This study proposes an innovative method for emotion recognition and behavior prediction in children with autism. By integrating emotional intelligence with behavioral analysis, the accuracy of emotion recognition and the stability of behavior prediction are improved. We use multi-layer perceptrons (MLPs) from deep learning models to identify emotions from multimodal data such as facial expressions, eye states, and head poses of children with autism. Combining these models with behavioral analysis models to predict dynamic behaviors, we propose an ensemble model that optimizes the final prediction results by training different emotion recognition models (such as random forests, gradient boosting trees, and neural networks) and integrating their outputs. Experimental results show that the proposed model achieves high precision in emotion recognition and behavior prediction tasks and can effectively analyze the emotional stability of children with autism. This method provides a new technical approach for intelligent perception and behavioral intervention in children with autism and demonstrates the potential of affective computing and behavioral analysis in practical applications.

BridgeVLA: Input-Output Alignment for Efficient 3D Manipulation Learning with Vision-Language Models

ArXiv.org · 2025-06-09

Recently, leveraging pre-trained vision-language models (VLMs) for building vision-language-action (VLA) models has emerged as a promising approach to effective robot manipulation learning. However, only few methods incorporate 3D signals into VLMs for action prediction, and they do not fully leverage the spatial structure inherent in 3D data, leading to low sample efficiency. In this paper, we introduce BridgeVLA, a novel 3D VLA model that (1) projects 3D inputs to multiple 2D images, ensuring input alignment with the VLM backbone, and (2) utilizes 2D heatmaps for action prediction, unifying the input and output spaces within a consistent 2D image space. In addition, we propose a scalable pre-training method that equips the VLM backbone with the capability to predict 2D heatmaps before downstream policy learning. Extensive experiments show the proposed method is able to learn 3D manipulation efficiently and effectively. BridgeVLA outperforms state-of-the-art baseline methods across three simulation benchmarks. In RLBench, it improves the average success rate from 81.4% to 88.2%. In COLOSSEUM, it demonstrates significantly better performance in challenging generalization settings, boosting the average success rate from 56.7% to 64.0%. In GemBench, it surpasses all the comparing baseline methods in terms of average success rate. In real-robot experiments, BridgeVLA outperforms a state-of-the-art baseline method by 32% on average. It generalizes robustly in multiple out-of-distribution settings, including visual disturbances and unseen instructions. Remarkably, it is able to achieve a success rate of 96.8% on 10+ tasks with only 3 trajectories per task, highlighting its extraordinary sample efficiency. Project Website:https://bridgevla.github.io/