Comparing Fine-Tuning, Zero and Few-Shot Strategies with Large Language Models in Hate Speech Detection in English

Large Language Models(LLMs)are increasingly demonstrating their ability to understand natural language and solve complex tasks,especially through text generation.One of the relevant capabilities is contextual learning,which involves the ability to receive instructions in natural language or task demonstrations to generate expected outputs for test instances without the need for additional training or gradient updates.In recent years,the popularity of social networking has provided a medium through which some users can engage in offensive and harmful online behavior.In this study,we investigate the ability of different LLMs,ranging from zero-shot and few-shot learning to fine-tuning.Our experiments show that LLMs can identify sexist and hateful online texts using zero-shot and few-shot approaches through information retrieval.Furthermore,it is found that the encoder-decoder model called Zephyr achieves the best results with the fine-tuning approach,scoring 86.811%on the Explainable Detection of Online Sexism(EDOS)test-set and 57.453%on the Multilingual Detection of Hate Speech Against Immigrants and Women in Twitter(HatEval)test-set.Finally,it is confirmed that the evaluated models perform well in hate text detection,as they beat the best result in the HatEval task leaderboard.The error analysis shows that contextual learning had difficulty distinguishing between types of hate speech and figurative language.However,the fine-tuned approach tends to produce many false positives.

Fine-tuning electronic structure of N-doped graphitic carbon-supported Co-and Fe-incorporated Mo_(2)C to achieve ultrahigh electrochemical water oxidation activity

Mo_(2)C is an excellent electrocatalyst for hydrogen evolution reaction(HER).However,Mo_(2)C is a poor electrocatalyst for oxygen evolution reaction(OER).Herein,two different elements,namely Co and Fe,are incorporated in Mo_(2)C that,therefore,has a finely tuned electronic structure,which is not achievable by incorporation of any one of the metals.Consequently,the resulting electrocatalyst Co_(0.8)Fe_(0.2)-Mo_(2)C-80 displayed excellent OER catalytic performance,which is evidenced by a low overpotential of 214.0(and 246.5)mV to attain a current density of 10(and 50)mA cm^(-2),an ultralow Tafel slope of 38.4 mV dec^(-1),and longterm stability in alkaline medium.Theoretical data demonstrates that Co_(0.8)Fe_(0.2)-Mo_(2)C-80 requires the lowest overpotential(1.00 V)for OER and Co centers to be the active sites.The ultrahigh catalytic performance of the electrocatalyst is attributed to the excellent intrinsic catalytic activity due to high Brunauer-Emmett-Teller specific surface area,large electrochemically active surface area,small Tafel slope,and low chargetransfer resistance.

Optimizing Enterprise Conversational AI: Accelerating Response Accuracy with Custom Dataset Fine-Tuning

As the realm of enterprise-level conversational AI continues to evolve, it becomes evident that while generalized Large Language Models (LLMs) like GPT-3.5 bring remarkable capabilities, they also bring forth formidable challenges. These models, honed on vast and diverse datasets, have undoubtedly pushed the boundaries of natural language understanding and generation. However, they often stumble when faced with the intricate demands of nuanced enterprise applications. This research advocates for a strategic paradigm shift, urging enterprises to embrace a fine-tuning approach as a means to optimize conversational AI. While generalized LLMs are linguistic marvels, their inability to cater to the specific needs of businesses across various industries poses a critical challenge. This strategic shift involves empowering enterprises to seamlessly integrate their own datasets into LLMs, a process that extends beyond linguistic enhancement. The core concept of this approach centers on customization, enabling businesses to fine-tune the AI’s functionality to fit precisely within their unique business landscapes. By immersing the LLM in industry-specific documents, customer interaction records, internal reports, and regulatory guidelines, the AI transcends its generic capabilities to become a sophisticated conversational partner aligned with the intricacies of the enterprise’s domain. The transformative potential of this fine-tuning approach cannot be overstated. It enables a transition from a universal AI solution to a highly customizable tool. The AI evolves from being a linguistic powerhouse to a contextually aware, industry-savvy assistant. As a result, it not only responds with linguistic accuracy but also with depth, relevance, and resonance, significantly elevating user experiences and operational efficiency. In the subsequent sections, this paper delves into the intricacies of fine-tuning, exploring the multifaceted challenges and abundant opportunities it presents. It addresses the technical intricacies of data integration, ethical considerations surrounding data usage, and the broader implications for the future of enterprise AI. The journey embarked upon in this research holds the potential to redefine the role of conversational AI in enterprises, ushering in an era where AI becomes a dynamic, deeply relevant, and highly effective tool, empowering businesses to excel in an ever-evolving digital landscape.

Genome-wide investigation to assess copy number variants in the Italian local chicken population

Background Copy number variants(CNV)hold significant functional and evolutionary importance.Numerous ongoing CNV studies aim to elucidate the etiology of human diseases and gain insights into the population structure of livestock.High-density chips have enabled the detection of CNV with increased resolution,leading to the identification of even small CNV.This study aimed to identify CNV in local Italian chicken breeds and investigate their distribution across the genome.Results Copy number variants were mainly distributed across the first six chromosomes and primarily associated with loss type CNV.The majority of CNV in the investigated breeds were of types 0 and 1,and the minimum length of CNV was significantly larger than that reported in previous studies.Interestingly,a high proportion of the length of chromosome 16 was covered by copy number variation regions(CNVR),with the major histocompatibility complex being the likely cause.Among the genes identified within CNVR,only those present in at least five animals across breeds(n=95)were discussed to reduce the focus on redundant CNV.Some of these genes have been associated to functional traits in chickens.Notably,several CNVR on different chromosomes harbor genes related to muscle development,tissue-specific biological processes,heat stress resistance,and immune response.Quantitative trait loci(QTL)were also analyzed to investigate potential overlapping with the identified CNVR:54 out of the 95 gene-containing regions overlapped with 428 QTL associated to body weight and size,carcass characteristics,egg production,egg components,fat deposition,and feed intake.Conclusions The genomic phenomena reported in this study that can cause changes in the distribution of CNV within the genome over time and the comparison of these differences in CNVR of the local chicken breeds could help in preserving these genetic resources.

Present situation and prospect of immunotherapy for unresectable locally advanced esophageal cancer during peri-radiotherapy

Four major studies(Checkmate577,Keynote-590,Checkmate649 and Attraction-4)of locally advanced esophageal cancer published in 2020 have established the importance of immunotherapy,represented by anti-programmed death protein(PD)-1 in postoperative adjuvant treatment and advanced first-line treatment of locally advanced or advanced esophageal cancer and esophagogastric junction cancer,from the aspects of proof of concept,long-term survival,overall survival rate and progression-free survival.For unresectable or inoperable nonmetastatic esophageal cancer,concurrent radiotherapy and chemotherapy is the standard treatment recommended by various guidelines.Because its curative effect is still not ideal,it is necessary to explore radical radiotherapy and chemotherapy in the future,and it is considered to be promising to combine them with immunotherapeutic drugs such as anti-PD-1.This paper mainly discusses how to combine radical concurrent radiotherapy and chemotherapy with immunotherapy for unresectable local advanced esophageal cancer.

Scheme Based on Multi-Level Patch Attention and Lesion Localization for Diabetic Retinopathy Grading

Early screening of diabetes retinopathy(DR)plays an important role in preventing irreversible blindness.Existing research has failed to fully explore effective DR lesion information in fundus maps.Besides,traditional attention schemes have not considered the impact of lesion type differences on grading,resulting in unreasonable extraction of important lesion features.Therefore,this paper proposes a DR diagnosis scheme that integrates a multi-level patch attention generator(MPAG)and a lesion localization module(LLM).Firstly,MPAGis used to predict patches of different sizes and generate a weighted attention map based on the prediction score and the types of lesions contained in the patches,fully considering the impact of lesion type differences on grading,solving the problem that the attention maps of lesions cannot be further refined and then adapted to the final DR diagnosis task.Secondly,the LLM generates a global attention map based on localization.Finally,the weighted attention map and global attention map are weighted with the fundus map to fully explore effective DR lesion information and increase the attention of the classification network to lesion details.This paper demonstrates the effectiveness of the proposed method through extensive experiments on the public DDR dataset,obtaining an accuracy of 0.8064.

Deep brain implantable microelectrode arrays for detection and functional localization of the subthalamic nucleus in rats with Parkinson’s disease

The subthalamic nucleus(STN)is considered the best target for deep brain stimulation treatments of Parkinson’s disease(PD).It is difficult to localize the STN due to its small size and deep location.Multichannel microelectrode arrays(MEAs)can rapidly and precisely locate the STN,which is important for precise stimulation.In this paper,16-channel MEAs modified with multiwalled carbon nanotube/poly(3,4-ethylenedioxythiophene):poly(styrene sulfonate)(MWCNT/PEDOT:PSS)nanocomposites were designed and fabricated,and the accurate and rapid identification of the STN in PD rats was performed using detection sites distributed at different brain depths.These results showed that nuclei in 6-hydroxydopamine hydrobromide(6-OHDA)-lesioned brains discharged more intensely than those in unlesioned brains.In addition,the MEA simultaneously acquired neural signals from both the STN and the upper or lower boundary nuclei of the STN.Moreover,higher values of spike firing rate,spike amplitude,local field potential(LFP)power,and beta oscillations were detected in the STN of the 6-OHDA-lesioned brain,and may therefore be biomarkers of STN localization.Compared with the STNs of unlesioned brains,the power spectral density of spikes and LFPs synchronously decreased in the delta band and increased in the beta band of 6-OHDA-lesioned brains.This may be a cause of sleep and motor disorders associated with PD.Overall,this work describes a new cellular-level localization and detection method and provides a tool for future studies of deep brain nuclei.

A study of the effect of local scour on the flow field near the spur dike

The flow field near a spur dike such as down flow and horseshoe vortex system(HVS)are susceptible to the topographic changes in the local scouring process,resulting in variation of the sediment transport with time.In this study,large eddy simulations with fixed-bed at different scouring stages were conducted to investigate the changes in flow field.The results imply that the bed deformation leads to an increase in flow rate per unit area,which represent the capability of sediment transportation by water,in the scour hole.Moreover,the intensity of turbulent kinetic energy and bimodal motion near the sand bed induced by the HVS were also varied.However,the peak moments between the two sediment transport mechanisms were different.Hence,understanding the complex feedback mechanism between topography and flow field is essential for the local scour problem.

Numerical study on local scour characteristics around submarine pipelines in the Yellow River Delta silty sandy soil under waves and currents

Due to their high reliability and cost-efficiency,submarine pipelines are widely used in offshore oil and gas resource engineering.Due to the interaction of waves,currents,seabed,and pipeline structures,the soil around submarine pipelines is prone to local scour,severely affecting their operational safety.With the Yellow River Delta as the research area and based on the renormalized group(RNG)k-εturbulence model and Stokes fifth-order wave theory,this study solves the Navier-Stokes(N-S)equation using the finite difference method.The volume of fluid(VOF)method is used to describe the fluid-free surface,and a threedimensional numerical model of currents and waves-submarine pipeline-silty sandy seabed is established.The rationality of the numerical model is verified using a self-built waveflow flume.On this basis,in this study,the local scour development and characteristics of submarine pipelines in the Yellow River Delta silty sandy seabed in the prototype environment are explored and the influence of the presence of pipelines on hydrodynamic features such as surrounding flow field,shear stress,and turbulence intensity is analyzed.The results indicate that(1)local scour around submarine pipelines can be divided into three stages:rapid scour,slow scour,and stable scour.The maximum scour depth occurs directly below the pipeline,and the shape of the scour pits is asymmetric.(2)As the water depth decreases and the pipeline suspension height increases,the scour becomes more intense.(3)When currents go through a pipeline,a clear stagnation point is formed in front of the pipeline,and the flow velocity is positively correlated with the depth of scour.This study can provide a valuable reference for the protection of submarine pipelines in this area.

Local dose-dense chemotherapy for triple-negative breast cancer via minimally invasive implantation of 3D printed devices

Dose-dense chemotherapy is the preferred first-line therapy for triple-negative breast cancer(TNBC),a highly aggressive disease with a poor prognosis.This treatment uses the same drug doses as conventional chemotherapy but with shorter dosing intervals,allowing for promising clinical outcomes with intensive treatment.However,the frequent systemic administration used for this treatment results in systemic toxicity and low patient compliance,limiting therapeutic efficacy and clinical benefit.Here,we report local dose-dense chemotherapy to treat TNBC by implanting 3D printed devices with timeprogrammed pulsatile release profiles.The implantable device can control the time between drug releases based on its internal microstructure design,which can be used to control dose density.The device is made of biodegradable materials for clinical convenience and designed for minimally invasive implantation via a trocar.Dose density variation of local chemotherapy using programmable release enhances anti-cancer effects in vitro and in vivo.Under the same dose density conditions,device-based chemotherapy shows a higher anticancer effect and less toxic response than intratumoral injection.We demonstrate local chemotherapy utilizing the implantable device that simulates the drug dose,number of releases,and treatment duration of the dose-dense AC(doxorubicin and cyclophosphamide)regimen preferred for TNBC treatment.Dose density modulation inhibits tumor growth,metastasis,and the expression of drug resistance-related proteins,including p-glycoprotein and breast cancer resistance protein.To the best of our knowledge,local dose-dense chemotherapy has not been reported,and our strategy can be expected to be utilized as a novel alternative to conventional therapies and improve anti-cancer efficiency.

Cooperative Anti-Jamming and Interference Mitigation for UAV Networks: A Local Altruistic Game Approach

To improve the anti-jamming and interference mitigation ability of the UAV-aided communication systems, this paper investigates the channel selection optimization problem in face of both internal mutual interference and external malicious jamming. A cooperative anti-jamming and interference mitigation method based on local altruistic is proposed to optimize UAVs’ channel selection. Specifically, a Stackelberg game is modeled to formulate the confrontation relationship between UAVs and the jammer. A local altruistic game is modeled with each UAV considering the utilities of both itself and other UAVs. A distributed cooperative anti-jamming and interference mitigation algorithm is proposed to obtain the Stackelberg equilibrium. Finally, the convergence of the proposed algorithm and the impact of the transmission power on the system loss value are analyzed, and the anti-jamming performance of the proposed algorithm can be improved by around 64% compared with the existing algorithms.

Enhancing Fire Detection Performance Based on Fine-Tuned YOLOv10

In recent years,early detection and warning of fires have posed a significant challenge to environmental protection and human safety.Deep learning models such as Faster R-CNN(Faster Region based Convolutional Neural Network),YOLO(You Only Look Once),and their variants have demonstrated superiority in quickly detecting objects from images and videos,creating new opportunities to enhance automatic and efficient fire detection.The YOLO model,especially newer versions like YOLOv10,stands out for its fast processing capability,making it suitable for low-latency applications.However,when applied to real-world datasets,the accuracy of fire prediction is still not high.This study improves the accuracy of YOLOv10 for real-time applications through model fine-tuning techniques and data augmentation.The core work of the research involves creating a diverse fire image dataset specifically suited for fire detection applications in buildings and factories,freezing the initial layers of the model to retain general features learned from the dataset by applying the Squeeze and Excitation attention mechanism and employing the Stochastic Gradient Descent(SGD)with a momentum optimization algorithm to enhance accuracy while ensuring real-time fire detection.Experimental results demonstrate the effectiveness of the proposed fire prediction approach,where the YOLOv10 small model exhibits the best balance compared to other YOLO family models such as nano,medium,and balanced.Additionally,the study provides an experimental evaluation to highlight the effectiveness of model fine-tuning compared to the YOLOv10 baseline,YOLOv8 and Faster R-CNN based on two criteria:accuracy and prediction time.

Assessment of earthquake location uncertainties for the design of local seismic networks

The ability to estimate earthquake source locations,along with the appraisal of relevant uncertainties,is paramount in monitoring both natural and human-induced micro-seismicity.For this purpose,a monitoring network must be designed to minimize the location errors introduced by geometrically unbalanced networks.In this study,we first review different sources of errors relevant to the localization of seismic events,how they propagate through localization algorithms,and their impact on outcomes.We then propose a quantitative method,based on a Monte Carlo approach,to estimate the uncertainty in earthquake locations that is suited to the design,optimization,and assessment of the performance of a local seismic monitoring network.To illustrate the performance of the proposed approach,we analyzed the distribution of the localization uncertainties and their related dispersion for a highly dense grid of theoretical hypocenters in both the horizontal and vertical directions using an actual monitoring network layout.The results expand,quantitatively,the qualitative indications derived from purely geometrical parameters(azimuthal gap(AG))and classical detectability maps.The proposed method enables the systematic design,optimization,and evaluation of local seismic monitoring networks,enhancing monitoring accuracy in areas proximal to hydrocarbon production,geothermal fields,underground natural gas storage,and other subsurface activities.This approach aids in the accurate estimation of earthquake source locations and their associated uncertainties,which are crucial for assessing and mitigating seismic risks,thereby enabling the implementation of proactive measures to minimize potential hazards.From an operational perspective,reliably estimating location accuracy is crucial for evaluating the position of seismogenic sources and assessing possible links between well activities and the onset of seismicity.

SNR and RSSI Based an Optimized Machine Learning Based Indoor Localization Approach:Multistory Round Building Scenario over LoRa Network

In situations when the precise position of a machine is unknown,localization becomes crucial.This research focuses on improving the position prediction accuracy over long-range(LoRa)network using an optimized machine learning-based technique.In order to increase the prediction accuracy of the reference point position on the data collected using the fingerprinting method over LoRa technology,this study proposed an optimized machine learning(ML)based algorithm.Received signal strength indicator(RSSI)data from the sensors at different positions was first gathered via an experiment through the LoRa network in a multistory round layout building.The noise factor is also taken into account,and the signal-to-noise ratio(SNR)value is recorded for every RSSI measurement.This study concludes the examination of reference point accuracy with the modified KNN method(MKNN).MKNN was created to more precisely anticipate the position of the reference point.The findings showed that MKNN outperformed other algorithms in terms of accuracy and complexity.

Survey of Indoor Localization Based on Deep Learning

This study comprehensively examines the current state of deep learning (DL) usage in indoor positioning.It emphasizes the significance and efficiency of convolutional neural networks (CNNs) and recurrent neuralnetworks (RNNs). Unlike prior studies focused on single sensor modalities like Wi-Fi or Bluetooth, this researchexplores the integration of multiple sensor modalities (e.g.,Wi-Fi, Bluetooth, Ultra-Wideband, ZigBee) to expandindoor localization methods, particularly in obstructed environments. It addresses the challenge of precise objectlocalization, introducing a novel hybrid DL approach using received signal information (RSI), Received SignalStrength (RSS), and Channel State Information (CSI) data to enhance accuracy and stability. Moreover, thestudy introduces a device-free indoor localization algorithm, offering a significant advancement with potentialobject or individual tracking applications. It recognizes the increasing importance of indoor positioning forlocation-based services. It anticipates future developments while acknowledging challenges such as multipathinterference, noise, data standardization, and scarcity of labeled data. This research contributes significantly toindoor localization technology, offering adaptability, device independence, and multifaceted DL-based solutionsfor real-world challenges and future advancements. Thus, the proposed work addresses challenges in objectlocalization precision and introduces a novel hybrid deep learning approach, contributing to advancing locationcentricservices.While deep learning-based indoor localization techniques have improved accuracy, challenges likedata noise, standardization, and availability of training data persist. However, ongoing developments are expectedto enhance indoor positioning systems to meet real-world demands.

Seropositivity and Other Determinants Associated with Toxoplasmosis in Local Buffalo in Iraq

A cross-sectional study was conducted to determine the Seropositivity of Toxoplasma gondii in water buffalo in three Iraqi governorates (Baghdad, Dhi Qar, and Maysan) and to estimate the risk aspects related to infestation throughout the period from January to December 2019. A total of 430 serum samples were inspected with a commercial ELISA (Enzyme linked immunosorbent assay) kit. Indirect multi-species kit. The overall Seropositivity of T. gondii in the examined local buffalo was 7.4%, and the highest rate (9.3%) was in Baghdad Governorate. A multivariate regression analysis revealed that adult buffalo (OR = 7.10;95% CI: 0.87-57.68;P = 0.067) and young herds (OR = 8.42;95% CI: 1.07-66.02;P = 0.043) were more subject to infestation from young buffalo and large herds. Furthermore, the hazard of toxoplasmosis was increased in winter especially among animals in contact with cats. It is therefore requisite to determine risk aspects to evaluate which mitigation, control, prevention and procedures should be carried out to diminish, control and prevent infestation with T. gondii and its propagation.

Improved PSO-Extreme Learning Machine Algorithm for Indoor Localization

Wi Fi and fingerprinting localization method have been a hot topic in indoor positioning because of their universality and location-related features.The basic assumption of fingerprinting localization is that the received signal strength indication(RSSI)distance is accord with the location distance.Therefore,how to efficiently match the current RSSI of the user with the RSSI in the fingerprint database is the key to achieve high-accuracy localization.In this paper,a particle swarm optimization-extreme learning machine(PSO-ELM)algorithm is proposed on the basis of the original fingerprinting localization.Firstly,we collect the RSSI of the experimental area to construct the fingerprint database,and the ELM algorithm is applied to the online stages to determine the corresponding relation between the location of the terminal and the RSSI it receives.Secondly,PSO algorithm is used to improve the bias and weight of ELM neural network,and the global optimal results are obtained.Finally,extensive simulation results are presented.It is shown that the proposed algorithm can effectively reduce mean error of localization and improve positioning accuracy when compared with K-Nearest Neighbor(KNN),Kmeans and Back-propagation(BP)algorithms.

Phase induced localization transition

Localization phenomenon is an important research field in condensed matter physics.However,due to the complexity and subtlety of disordered systems,new localization phenomena always emerge unexpectedly.For example,it is generally believed that the phase of the hopping term does not affect the localization properties of the system,so the calculation of the phase is often ignored in the study of localization.Here,we introduce a quasiperiodic model and demonstrate that the phase change of the hopping term can significantly alter the localization properties of the system through detailed numerical simulations,such as the inverse participation ratio and multifractal analysis.This phase-induced localization transition provides valuable information for the study of localization physics.

Navigating the pathways:TAR-DNA-binding-protein-43 aggregation,axonal transport,and local synthesis in amyotrophic lateral sclerosis pathology

Neurons are highly polarized cells with axons reaching over a meter long in adult humans.To survive and maintain their proper function,neurons depend on specific mechanisms that regulate spatiotemporal signaling and metabolic events,which need to be carried out at the right place,time,and intensity.Such mechanisms include axonal transport,local synthesis,and liquid-liquid phase separations.Alterations and malfunctions in these processes are correlated to neurodegenerative diseases such as amyotrophic lateral sclerosis(ALS).

Ab initio density response and local field factor of warm dense hydrogen

We present quasi-exact ab initio path integral Monte Carlo(PIMC)results for the partial static density responses and local field factors of hydrogen in the warm dense matter regime,from solid density conditions to the strongly compressed case.The full dynamic treatment of electrons and protons on the same footing allows us to rigorously quantify both electronic and ionic exchange–correlation effects in the system,and to compare the results with those of earlier incomplete models such as the archetypal uniform electron gas or electrons in a fixed ion snapshot potential that do not take into account the interplay between the two constituents.The full electronic density response is highly sensitive to electronic localization around the ions,and our results constitute unambiguous predictions for upcoming X-ray Thomson scattering experiments with hydrogen jets and fusion plasmas.All PIMC results are made freely available and can be used directly for a gamut of applications,including inertial confinement fusion calculations and the modeling of dense astrophysical objects.Moreover,they constitute invaluable benchmark data for approximate but computationally less demanding approaches such as density functional theory or PIMC within the fixed-node approximation.