Abstract

We present LayForge, a two-track lay sum- mary generation system developed for the Bio- LaySumm2025 shared task(Xiao et al., 2025) . Task 1.1 addresses the lay summarization using only the internal content of the article, while Task 1.2 augments this process with do- main knowledge such as biomedical definitions and concept explanations. BioLaySumm em- ploys a modular architecture that leverages large language models (LLMs), a BioBERT- based named entity recognizer (NER), and the UMLS(Bodenreider, 2004) knowledge base to create readable, informative, and faithful lay summaries. Our system shows strong perfor- mance on both tasks when evaluated on the PLOS and elife subset(Goldsack et al., 2022), particularly in readability and factuality metrics. The architecture illustrates how modularity and domain adaptation can be effectively combined for accessible biomedical communication.

Abstract

This paper presents the development of a real-time object detection system using Frequency Modulated Continuous Wave(FMCW) millimeter-wave radar signals and the PYNQ-ZU FPGA Board, which is widely used in (advanced driving assistance system)ADAS and robotic applications. A hardware FPGA platform serves as a valid embedded architecture for the purpose of validating object detection and recognition applications. We used our experiment’s point cloud images to apply different machine learning models to detect these objects. Using a top-view (TV) filter to convert 3D point cloud images into 2D representations made object detection more accurate. Following the use of filtration techniques, we extracted features from the filtered 2D image using the VGG 16 model. We then assessed four machine learning models for object detection and found that the Support Vector Machine (SVM) model and Logistic Regression (LR) had better results, obtaining an accuracy of 97%. Our proposed work uses mmWave radar, Top-View Filter, VGG 16 Model, and LR to highly increase object detection accuracy over existing methods.

Abstract

Robotic inspection of transmission lines presents significant challenges due to the complexity of navigating along the wires. Existing systems often rely on either flight modes for visual inspection or articulated crawling mechanisms for contact-based inspection. However, these approaches face limitations in effectively bypassing in-line obstacles or pylons, which are common in transmission line environments. This letter presents WireFlie, a novel hybrid robotic system that integrates rolling and flight modes to overcome these challenges. The system consists of a pair of underactuated arms mounted on a drone platform, designed for secure, collision-free locking and detaching, enabling seamless transitions between locomotion modes. WireFlie supports both single-arm and dual-arm rolling, allowing it to bypass in-line obstacles such as Stockbridge dampers, dual spacers, and sleeves, and to overcome larger obstacles like pylons using flight. Additionally, we propose a high-level controller for autonomous locking, detaching, and obstacle avoidance. Experiments are conducted on a custom-made setup that closely resembles a transmission wire. We evaluate both the design and control aspects of our system, with results including kinematic analysis, wire detection, autonomous locking and the corresponding trajectory, and obstacle detection and avoidance strategies. This research contributes to the field of robotic infrastructure inspection by merging aerial and wire-based locomotion, providing efficient and autonomous monitoring of power lines.

Abstract

Emotion detection in text plays a very crucial role in NLP applications such as sentiment anal- ysis and feedback analysis. This study tackles two tasks: multi-label emotion detection, where the goal is to classify text based on six emotions (joy, sadness, fear, anger, surprise, and disgust) in a multilingual setting, and emotion intensity prediction, which assigns an ordinal intensity score to each of the above-given emotions. Using the BRIGHTER dataset, a multilingual corpus spanning 28 languages, the paper ad- dresses issues like class imbalances by treating each emotion as an independent binary classifi- cation problem. The paper first explores strate- gies such as static embeddings such as GloVe with logistic regression classifiers on top of it. To capture contextual nuances more effectively, we fine-tune transformer based models, such as BERT and RoBERTa. Our approach demon- strates the benefits of fine-tuning for improved emotion prediction, while also highlighting the challenges of multilingual and multi-label clas- sification.

Abstract

Machine Translation (MT) is an essential tool for communication among people across different cultures, yet Named Entity (NE) translation remains a major challenge due to its rarity in occurrence and ambiguity. Traditional approaches, like using lexicons or parallel corpora, often fail to generalize to unseen entities and, hence, do not perform well. To address this, we create a silver dataset using the Google Translate API and fine-tune the facebook/nllb200-distilled-600M model with LoRA (LowRank Adaptation) to enhance translation accuracy while also maintaining efficient memory use. Evaluated with metrics such as BLEU, COMET, and M-ETA, our results show that fine-tuning a specialized MT model improves NE translation without having to rely on largescale general-purpose models.

Abstract

Many real-world systems can be modeled as dynamic graphs, where nodes and edges evolve over time, requiring specialized models to capture their evolving dynamics in risk-sensitive applications effectively. Graph neural networks (GNNs) for temporal graphs are one such category of specialized models. For the first time, our approach integrates a reject option strategy within the framework of GNNs for continuous time dynamic graphs (CTDGs). This allows the model to strategically abstain from making predictions when the uncertainty is high and confidence is low, thus minimizing the risk of critical misclassification and enhancing the results and reliability. We propose a coverage-based abstention prediction model to implement the reject option that maximizes prediction within a specified coverage. It improves the prediction score for link prediction and node classification tasks. Temporal GNNs deal with extremely skewed datasets for the next state prediction or node classification task. In the case of class imbalance, our method can be further tuned to provide a higher weight to the minority class. Exhaustive experiments are presented on four datasets for dynamic link prediction and two datasets for dynamic node classification tasks. This demonstrates the effectiveness of our approach in improving the reliability and area under the curve (AUC)/average precision (AP) scores for predictions in dynamic graph scenarios. The results highlight our model’s ability to efficiently handle the trade-offs between prediction confidence and coverage, making it a dependable solution for applications requiring high precision in dynamic and uncertain environments.

Abstract

Sewage Treatment Plants (STPs) are critical for public health and environmental sustainability. However, Traditional water quality monitoring in globally heavily relies on labour-intensive, lab-based methods, which pose significant challenges but in recent years, IoT-based sensors have emerged as a promising alternative. However, these sensors face challenges including sensor fouling and frequent manual maintenance, which compromise long-term accuracy and operational efficiency. This study presents an innovative IoT enabled Self-Cleaning Water Quality Monitoring System (SC-WQMS) to address these challenges through automation and precision. The proposed system integrates advanced features, including self-cleaning mechanisms. Key water quality parameter, such as Total Dissolved Solids (TDS) is analysed in real-time by using IoT-based sensors. The proposed IoT-enabled SC-WQMS demonstrates significantly better accuracy over traditional non-cleaning devices. Lab report data over a two-month period reveals that the SC-WQMS maintains an average accuracy of 95%-98.5% for TDS measurements, compared to 68%-75% accuracy observed in traditional IoT-based non cleaning devices due to issues like algae formation and particle buildup over the same period. This substantial difference highlights the SC-WQMS's effectiveness in ensuring reliable data transmission to a cloud server. By mitigating the challenges faced by traditional methods. The proposed SC-WQMS represents a scalable, efficient, and sustainable solution for modern water treatment facilities, addressing the operational challenges of traditional systems while promoting better resource management and environmental protection.

Abstract

We introduce Berman-intersection-dual Berman (BiD) codes. These are abelian codes of length $3^m$ that can be constructed using Kronecker products of a $3 times 3$ kernel matrix. BiD codes offer minimum distance close to that of Reed-Muller (RM) codes at practical blocklengths, and larger distance than RM codes asymptotically in the blocklength. Simulations of BiD codes of length $3^5=243$ in the erasure and Gaussian channels show that their block error rates under maximum-likelihood decoding are similar to, and sometimes better, than RM, RM-Polar, and CRC-aided Polar codes.

Abstract

Diffusion-based motion planners are becoming popular due to their well-established performance improvements, stemming from sample diversity and the ease of incorporating new constraints directly during inference. However, a primary limitation of the diffusion process is the requirement for a substantial number of denoising steps, especially when the denoising process is coupled with gradient-based guidance. In this paper, we introduce, diffusion in the parametric space of trajectories, where the parameters are represented as Bernstein coefficients. We show that this representation greatly improves the effectiveness of the cost function guidance and the inference speed. We also introduce a novel stitching algorithm that leverages the diversity in diffusion-generated trajectories to produce collision-free trajectories with just a single cost function-guided model. We demonstrate that our approaches outperform current SOTA diffusion-based motion planners for manipulators and provide an ablation study on key components.

Abstract

Software developers commonly use logging mechanisms to gather runtime data. Over the years, this information has been used for various purposes like debugging, behavioral analysis, system comprehension, etc. Frameworks such as log4j and Logback have helped standardize logging practices by incorporating Software Engineering principles. Logging interactions in Virtual Reality (VR) applications can help understand user behavior and assist with automated conformance checks. In this paper, we introduce VRSLOG, a generalized logging framework that can capture interactions across diverse VR scenes without any changes to the framework. We implement a prototype of the VRSLOG framework and demonstrate the generation of log files. Further, the log files are used to conduct conformance checks against a predefined expected sequence of events within the VR scene.