Recent progress in large language models (LLMs) has made knowledge-grounded reasoning increasingly practical, yet KG-based QA systems often pay a steep price in efficiency and transparency. In typical pipelines, symbolic paths are scored by neural encoders or repeatedly re-ranked by multiple LLM calls, which inflates latency and GPU cost and makes the decision process hard to audit. We introduce PathHD, an encoder-free framework for knowledge-graph reasoning that couples hyperdimensional computing (HDC) with a single LLM call per query. Given a query, PathHD represents relation paths as block-diagonal GHRR hypervectors, retrieves candidate paths using a calibrated blockwise cosine similarity with Top-K pruning, and then performs a one-shot LLM adjudication that outputs the final answer together with supporting, citeable paths. The design is enabled by three technical components: (i) an order-sensitive, non-commutative binding operator for composing multi-hop paths, (ii) a robust similarity calibration that stabilizes hypervector retrieval, and (iii) an adjudication stage that preserves interpretability while avoiding per-path LLM scoring. Across WebQSP, CWQ, and GrailQA, PathHD matches or improves Hits@1 compared to strong neural baselines while using only one LLM call per query, reduces end-to-end latency by $40-60\%$, and lowers GPU memory by $3-5\times$ due to encoder-free retrieval. Overall, the results suggest that carefully engineered HDC path representations can serve as an effective substrate for efficient and faithful KG-LLM reasoning, achieving a strong accuracy-efficiency-interpretability trade-off.
Graph-Structured Hyperdimensional Computing for Data-Efficient and Explainable Process-Structure-Property Prediction
School of Mechanical, Aerospace, and Manufacturing Engineering, University of Connecticut(康奈尔大学机械、航空航天与制造工程学院);U.S. Army Combat Capabilities Development Command, Ground Vehicle Systems Center(美国陆军战斗能力发展指挥部地面车辆系统中心)
CoT-MAE v2: Contextual Masked Auto-Encoder with Multi-view Modeling for Passage Retrieval
Institute of Information Engineering, Chinese Academy of Sciences(中国科学院信息工程研究所);School of Cyber Security, University of Chinese Academy of Sciences(中国科学院大学网络空间安全学院);Kuaishou Technology(快手科技)
SEA: Sentence Encoder Assembly for Video Retrieval by Textual Queries
Renmin University of China(中国人民大学)
通过文本查询来检索无标签视频,称为Ad-hoc Video Search(AVS),是多媒体数据管理和检索的核心主题。AVS的成功依赖于跨模态表示学习,该学习将查询语句和视频编码到公共空间中以进行语义相似度计算。受先前少数作品在结合多个句子编码器方面取得的初步成功的启发,本文通过开发一种新的通用方法来有效地利用不同的句子编码器,向前迈出了一步。所提出的方法,我们术语句子编码器组件(SEA)的新颖性,是双重的。首先,与仅使用单个公共空间的现有技术不同,SEA支持多个编码器特定的公共空间中的文本-视频匹配。这样的属性防止匹配被产生比其他编码器长得多的编码向量的特定编码器支配。其次,为了探索个体公共空间之间的互补性,我们提出了多空间多损失学习。在MSR-VTT、TRECVID AVS 2016-2019、TGIF和MSVD四个基准测试上的大量实验表明,SEA超越了最先进的水平。此外,SEA非常容易实现。所有这些都使SEA成为AVS的一个有吸引力的解决方案,并有希望通过收获新的句子编码器来不断推进任务。
MURAL: Multimodal, Multitask Retrieval Across Languages
Google Research(谷歌研究院)
图像-标题对和翻译对都提供了学习语言之间的深层表示和联系的方法。我们在MURAL(Multimodal,Multitask Representations Across Languages)中使用这两种类型的对,这是一种双重编码器,可以解决两个任务:1)图像-文本匹配和2)翻译对匹配。通过整合数十亿个翻译对,MURAL扩展了ALIGN(Jia et al. PMLR'21)--一种从18亿个嘈杂的图像-文本对中学习的最先进的双编码器。当使用相同的编码器时,MURAL的性能匹配或超过ALIGN的跨模态检索性能,在多个数据集上的资源丰富的语言。更重要的是,它大大提高了资源不足语言的性能,表明文本-文本学习可以克服这些语言缺乏图像标题示例的问题。例如,在维基百科的图像-文本数据集上,MURAL-base对8种资源不足的语言平均提高了zero-shot平均召回率8.1%,在微调时平均提高了6.8%。我们还表明,MURAL的文本表示集群不仅与家谱的连接,但也基于区域语言学,如巴尔干Sprachbund。
Recognizing Families through Images with Pretrained Encoder
Faculty of Information Technology, University of Science, VNU-HCM(越南国家大学胡志明市分校科学大学信息技术学院)
亲属关系验证和亲属关系检索是计算机视觉中的新兴任务。亲属关系验证的目的是确定两个人脸图像是否来自相关的人,而亲属关系检索是从图像库中检索可能与一个人相关的人脸图像的任务。他们引入了独特的挑战,因为隐藏的关系和功能,进行面部图像之间的固有特性。我们采用3种方法,FaceNet,Siamese VGG-Face以及FaceNet和VGG-Face模型的组合作为特征提取器,在2020年的Recognizing Family in The Wild比赛中获得了第9名的亲属关系验证和第5名的亲属关系检索。然后,我们使用StyleGAN 2作为另一个编码器进行了进一步的实验,结果没有任何改善。