Hologram Representation via Quadratic Phase Gaussian Splatting
本文提出了一种新的全息图表示方法CVQPG，通过使用二次相位函数替代传统的2D高斯表示，并增加可学习参数控制基底曲率，提高了全息重建的视觉质量。
0 citationsRead paper
Selected work
Representative Papers
Situated Action in Pre-Hospital Critical Care Dispatch: Identifying where and how Algorithmic Assistance might be useful in the daily work of specialist Emergency Medical Dispatchers
研究通过民族志沉浸和观察方法，分析急救医疗调度工作流程，旨在识别AI辅助在哪个环节最能有效减轻医务人员工作负担并改善临床结果。
0 citationsRead paper
Toward Cultural Alignment: Human-Centered Evaluation of Multimodal AI Stories Across Five African Communities
本文通过结合定量标注和定性讨论的方法，评估AI生成的多模态故事与五个非洲社区文化的一致性，并提出一种文化一致性的分类方法。
0 citationsRead paper
MirrorWorld: Taming Video Diffusion Models for Mirror Reflection Generation
This work addresses the common failure of video diffusion models to generate physically plausible mirror reflections, often resulting in semantic inconsistencies or spatial distortions due to neglecting the geometric and semantic coherence between scenes and their reflections. To overcome this, the authors propose a novel video inpainting framework specifically designed for mirror reflection synthesis, which decouples reflection generation into two complementary subtasks: “what to reflect” (semantic content) and “how to arrange it” (spatial layout). The approach leverages Semantic Relation Distillation (SRD) to transfer semantic correlations from a frozen vision foundation model and incorporates Geometric Transformation Alignment (GTA) to model the spatial transformation inherent in reflections. Evaluated on a newly established unified benchmark for video mirror reflection reconstruction, the method significantly outperforms existing image-level reflection generation and video inpainting baselines, achieving high-fidelity and spatially consistent mirror reflections.
0 citationsRead paper
Statistically Supported LLM Ingredient and Recipe Data Collection in Computational Nutrition
Existing nutritional databases are commonly plagued by incompleteness and inconsistency, and are designed primarily for human consultation, rendering them inadequate for the precise, structured data demands of computational nutrition. This work proposes an automated method for constructing high-quality ingredient-level nutritional data using large language models (LLMs). By treating LLM outputs as samples from a probability distribution, the approach operationalizes uncertainty and enables self-correction through repeated querying, robust statistical estimation, semantic consistency constraints, validation against nutritional logic invariants, and web-based evidence tracing. Evaluated on a test set of 30 ingredients, the method achieves a 98.4% exact-match accuracy for nutrient labels and reduces the median absolute percentage error in nutrient ratios from 31.9% to 10.1%, at an approximate API cost of one U.S. dollar per ingredient.
0 citationsRead paper
Recent publications
Latest Papers
Hologram Representation via Quadratic Phase Gaussian Splatting
本文提出了一种新的全息图表示方法CVQPG，通过使用二次相位函数替代传统的2D高斯表示，并增加可学习参数控制基底曲率，提高了全息重建的视觉质量。
0 citationsRead paper
Situated Action in Pre-Hospital Critical Care Dispatch: Identifying where and how Algorithmic Assistance might be useful in the daily work of specialist Emergency Medical Dispatchers
研究通过民族志沉浸和观察方法，分析急救医疗调度工作流程，旨在识别AI辅助在哪个环节最能有效减轻医务人员工作负担并改善临床结果。
0 citationsRead paper
Toward Cultural Alignment: Human-Centered Evaluation of Multimodal AI Stories Across Five African Communities
本文通过结合定量标注和定性讨论的方法，评估AI生成的多模态故事与五个非洲社区文化的一致性，并提出一种文化一致性的分类方法。
0 citationsRead paper
MirrorWorld: Taming Video Diffusion Models for Mirror Reflection Generation
This work addresses the common failure of video diffusion models to generate physically plausible mirror reflections, often resulting in semantic inconsistencies or spatial distortions due to neglecting the geometric and semantic coherence between scenes and their reflections. To overcome this, the authors propose a novel video inpainting framework specifically designed for mirror reflection synthesis, which decouples reflection generation into two complementary subtasks: “what to reflect” (semantic content) and “how to arrange it” (spatial layout). The approach leverages Semantic Relation Distillation (SRD) to transfer semantic correlations from a frozen vision foundation model and incorporates Geometric Transformation Alignment (GTA) to model the spatial transformation inherent in reflections. Evaluated on a newly established unified benchmark for video mirror reflection reconstruction, the method significantly outperforms existing image-level reflection generation and video inpainting baselines, achieving high-fidelity and spatially consistent mirror reflections.
0 citationsRead paper
Statistically Supported LLM Ingredient and Recipe Data Collection in Computational Nutrition
Existing nutritional databases are commonly plagued by incompleteness and inconsistency, and are designed primarily for human consultation, rendering them inadequate for the precise, structured data demands of computational nutrition. This work proposes an automated method for constructing high-quality ingredient-level nutritional data using large language models (LLMs). By treating LLM outputs as samples from a probability distribution, the approach operationalizes uncertainty and enables self-correction through repeated querying, robust statistical estimation, semantic consistency constraints, validation against nutritional logic invariants, and web-based evidence tracing. Evaluated on a test set of 30 ingredients, the method achieves a 98.4% exact-match accuracy for nutrient labels and reduces the median absolute percentage error in nutrient ratios from 31.9% to 10.1%, at an approximate API cost of one U.S. dollar per ingredient.
0 citationsRead paper