Reasoning with knowledge graphs (KGs) has primarily focused on triple-shaped facts. Recent advancements have been explored to enhance the semantics of these facts by incorporating more potent representations, such as hyper-relational However, approaches are limited atomic facts, which describe a single piece information. This paper extends beyond and delves into nested represented quoted triples where subjects objects themselves (e.g., ((BarackObama, holds_position, President), succeed_by,...

With the rapid growth of internet finance and booming financial lending, intelligent calling for debt collection in FinTech companies has driven increasing attention. Nowadays, widely used system is based on dialogue flow, namely configuring interaction flow with finite-state machine. In our scenario collection, completed contains more than one thousand interactive paths. All procedures are artificially specified, extremely high maintenance costs error-prone. To solve this problem, we...

A common approach for aligning language models to human preferences is first learn a reward model from preference data, and then use this update the model. We study two closely related problems that arise in approach. First, any monotone transformation of preserves ranking; there choice ``better'' than others? Second, we often wish align multiple properties: how should combine models? Using probabilistic interpretation alignment procedure, identify natural (the case of) rewards learned...

Despite the superb performance in many tasks, large language models (LLMs) bear risk of generating hallucination or even wrong answers when confronted with tasks that demand accuracy knowledge. The issue becomes more noticeable addressing logic queries require multiple reasoning steps. On other hand, knowledge graph (KG) based question answering methods are capable accurately identifying correct help graph, yet its could quickly deteriorate itself is sparse and incomplete. It remains a...

Developing agents that can follow multimodal instructions remains a fundamental challenge in robotics and AI. Although large-scale pre-training on unlabeled datasets (no language instruction) has enabled to learn diverse behaviors, these often struggle with following instructions. While augmenting the dataset instruction labels mitigate this issue, acquiring such high-quality annotations at scale is impractical. To address we frame problem as semi-supervised learning task introduce GROOT-2,...

Cross-view geo-localization identifies the locations of street-view images by matching them with geo-tagged satellite or OSM. However, most studies focus on image-to-image retrieval, fewer addressing text-guided a task vital for applications like pedestrian navigation and emergency response. In this work, we introduce novel cross-view natural language descriptions, which aims to retrieve corresponding OSM database based scene text. To support task, construct CVG-Text dataset collecting data...

Deep neural networks, empowered by pre-trained language models, have achieved remarkable results in natural understanding (NLU) tasks. However, their performances can drastically deteriorate when logical reasoning is needed. This because NLU principle depends on not only analogical reasoning, which deep networks are good at, but also reasoning. According to the dual-process theory, and respectively carried out System 1 2 human brain. Inspired we present a novel framework for called...

This paper presents a novel open-domain dialogue generation model emphasizing the differentiation of speakers in multi-turn conversations. Differing from prior work that solely relies on content conversation history to generate response, we argue capturing relative social relations among utterances (i.e., generated by either same speaker or different persons) benefits machine fine-grained context information improve coherence response. Given that, propose speaker-aware Parallel Hierarchical...

Context modeling plays a significant role in building multi-turn dialogue systems. In order to make full use of context information, systems can Incomplete Utterance Rewriting(IUR) methods simplify the into single-turn by merging current utterance and information self-contained utterance. However, previous approaches ignore intent consistency between original query rewritten query. The detection omitted or coreferred locations be further improved. this paper, we introduce contrastive...

Knowledge Graphs, such as Wikidata, comprise structural and textual knowledge in order to represent knowledge. For each of the two modalities dedicated approaches for graph embedding language models learn patterns that allow predicting novel Few have integrated learning inference with both these existing ones could only partially exploit interaction In our approach, we build on strong representations single use hypercomplex algebra both, (i), single-modality well as, (ii), between different...

The annotation for Chinese genealogy textual documents is helpful constructing knowledge graph, training effective machine learning models extraction, etc. However, this kind of difficult to annotate. primary reason that the texts are written in both classical and vernacular Chinese. These also contain numerous ancient characters usually without punctuation. Understanding requires sufficient expertise. When multiple users labeling same text, conflicts may occur. Existing tools inappropriate...

We study the problem of building a controller that can follow open-ended instructions in open-world environments. propose to reference videos as instructions, which offer expressive goal specifications while eliminating need for expensive text-gameplay annotations. A new learning framework is derived allow such instruction-following controllers from gameplay producing video instruction encoder induces structured space. implement our agent GROOT simple yet effective encoder-decoder...

利用多体量子纠缠度量，研究了与一维半无限波导耦合的三个两能级原子系统纠缠产生问题。考虑原子之间的距离、原子和波导之间的手性耦合、和波导有限界与原子系统的距离对量子纠缠的影响。由于镜子导致的量子反馈，使得量子系统的方程为延迟微分方程。通过求解延迟微分方程，得到原子系统的量子纠缠演化。分析发现和一维无限波导情形相比，耦合于半无限波导的原子系统中产生的量子纠缠能够持续更长时间，而且其动力学特性依赖于原子之间的距离、手性耦合强度和原子的位置。在不考虑延迟情形下，分析了手性耦合对量子纠缠的影响，发现手性耦合强度越大，量子纠缠到达最大值越快，而后更快地衰减到零。研究表明可以通过控制原子之间的距离、手性耦合强度和原子的位置来制备量子纠缠。研究对理解基于波导的多体量子纠缠产生有重要意义。