Personalizing large language model (LLM) responses in enterprise environments faces unique challenges: heterogeneous data sources, strict on-premise constraints, and unreliable LLM outputs. We present PersonaPlugin, a production-ready framework that transforms real-world telecom signals—call summaries, app usage, and location patterns—into structured personas for LLM personalization. Our system processes 1M+ behavioral records daily, entirely on-premise. The framework comprises three specialized engines, with the call extraction engine achieving >99% parsing success through cascaded fallback mechanisms. We conduct a comprehensive two-track evaluation: (1) general personalization across 800 scenarios, and (2) call-context personalization across 400 scenarios. Our LLM-as-a-Judge evaluation, validated against human judgments (κ=0.71), demonstrates that persona injection significantly improves personalization: +24.0% in P-Score and +26.7% in Adherence over baseline. Integrating call context with personas yields +11% additional improvement, validating the synergistic value of multi-source persona construction.

Time series classification (TSC) presents significant challenges in data analytics and plays an essential role in industries such as manufacturing, healthcare, and finance. Existing research has utilized sequence models, such as long short-term memory (LSTM) and transformers, to achieve high performance; however, these models fail to fully address the inherent challenges of capturing long-term dependencies in time series data. Recently, efforts have been made to overcome these limitations by processing time series data through conversion into images. One such method, the Gramian Angular Field (GAF), converts time series data into images that visually represent the relationship between time points, effectively capturing global trends that traditional sequence models often miss. However, single-modality approaches, which rely on only one representation of time series data, still face limitations in comprehensively capturing the diverse features of a time series. Additionally, existing multi-modality approaches may overlook the detailed features between time points. To address these issues, this article proposes a patch-level hybrid contrastive learning model that combines transformer with Vision Transformer (ViT). The proposed model converts a one-dimensional time series into GAF images and leverages contrastive learning to robustly learn fine-grained features. This enables the model to effectively capture both temporal relationships between time points and spatial relationships between patches, thereby enhancing its generalization capabilities. Experimental results demonstrate that the proposed model outperforms the existing methods on the UCR TSC dataset. This shows that the model can overcome the limitations of single-modality approaches by integrating the complex structural features of time series data through a multi-modality framework, ultimately leading to improved classification performance. This approach provides a new avenue for enhancing TSC and holds promise for applications across various industries.

Most studies on abstractive summarization are conducted in a supervised learning framework, aiming to generate a golden summary from the original document. In this process, the model focuses on portions of the document that closely resemble the golden summary to produce a coherent output. Consequently, current methodologies tend to achieve higher performance on extractive datasets compared to abstractive datasets, indicating diminished effectiveness on more abstracted content. To address this, our study proposes a methodology that maintains high effectiveness on abstractive datasets. Specifically, we introduce a multi-task learning approach that incorporates both salient and non-salient information during training. This is implemented by adding a contrastive objective to the fine-tuning phase of an encoder–decoder language model. Salient and non-salient parts are selected based on ROUGE-L F1 scores, and their relationships are learned through a triplet loss function. The proposed method is evaluated on five benchmark summarization datasets, including two extractive and three abstractive datasets. Experimental results demonstrate significant performance improvements on abstractive datasets, particularly those with high levels of abstraction, compared to existing abstractive summarization methods.

As the digital transformation of the manufacturing industry accelerates, the importance of data integration in manufacturing processes is increasingly emphasized. However, the use of heterogeneous equipment and diverse solutions in building processes hinders smooth data exchange, resulting in compatibility issues among equipment, systems, and data. To address these challenges, the Reference Architecture Model Industry 4.0 (RAMI 4.0) has been proposed as a reference model for Industry 4.0, utilizing the Asset Administration Shell (AAS) to ensure both vertical and horizontal interoperability. In this study, a robot manipulator was modeled using AAS, and the OPC UA protocol was applied to the modeled process elements. Additionally, a digital twin of the robot manipulator was built using MoveIt2 from higher-level systems, allowing for parameter-based control. This facilitates scenario-based integration with heterogeneous equipment such as PLCs, enabling the design of a flexible robot production system that can adapt to the tasks of heterogeneous equipment without human intervention.

Recent graph-based face clustering methods predict the connectivity of enormous edges, including false positive edges that link nodes with different classes. However, those false positive edges, which connect negative node pairs, have the risk of integration of different clusters when their connectivity is incorrectly estimated. This paper proposes a novel face clustering method to address this problem. The proposed clustering method employs density-based clustering, which maintains edges that have higher density. For this purpose, we propose a reliable density estimation algorithm based on local connectivity between K nearest neighbors (KNN). We effectively exclude negative pairs from the KNN graph based on the reliable density while maintaining sufficient positive pairs. Furthermore, we develop a pairwise connectivity estimation network to predict the connectivity of the selected edges. Experimental results demonstrate that the proposed clustering method significantly outperforms the state-of-the-art clustering methods on large-scale face clustering datasets and fashion image clustering datasets. Our code is available at https://github.com/illian01/LCE-PCENet

Query auto completion (QAC) is the task of predicting a search engine user’s final query from their intermediate, incomplete query. In this paper, we extend QAC to the streaming voice search setting, where automatic speech recognition systems produce intermediate transcriptions as users speak. Naively applying existing methods fails because the intermediate transcriptions often don’t form prefixes or even substrings of the final transcription. To address this issue, we propose to condition QAC approaches on intermediate transcriptions to complete voice queries. We evaluate our models on a speech-enabled smart television with real-life voice search traffic, finding that this ASR-aware conditioning improves the completion quality. Our best method obtains an 18% relative improvement in mean reciprocal rank over previous methods.

Flexibility has become an important characteristic of today's manufacturing industry. Unfortunately, today's standard of using a fixed (linear) process plan is not flexible enough to adapt to different manufacturing system conditions aiming at increasing the system's efficiency. One means to overcome such a limitation is to prepare pre-planned alternatives in the process plan. If this can be easily done and implemented, the most appropriate plan can be selected in real time from all possible alternatives according to the conditions of the shop floor. The objective of the paper is to demonstrate this concept in detail and to discuss the influence of process plan alternatives in a dynamic shop floor environment at equipment level. In order to verify whether the presence of alternatives in process plans increases the efficiency of the manufacturing system, a simulation will be carried out using some example parts currently being manufactured in Penn State's Factory for Advanced Manufacturing Education (FAME).