Persuasion has long been considered a unique human skill, rooted in emotion, context and intent. However, with rapid advances in communication technologies, particularly in Artificial Intelligence (AI), machines are beginning to replicate persuasive behaviour in more advanced ways. As AI becomes increasingly integrated into our daily lives, through virtual assistants and chatbots in areas ranging from customer service to healthcare consultations, a crucial question emerges: can language models not only inform but also persuade? Despite their widespread adoption, little is known about how effectively these models can influence human thinking, or how different prompting methods shape their responses. This paper presents an evaluation of state-of-the-art Large Language Models (LLMs) and Vision-Language Models (VLMs) in the domain of conversational AI. We assess seven LLMs and five VLMs by generating conversations based on gold-standard, human-written responses drawn from three persuasion datasets. Each model is evaluated using four distinct prompting techniques: zero-shot, one-shot, few-shot, and chain-of-thought prompting. Our goal is to analyse how these methods influence the models' capabilities in generating persuasive responses. Both automatic and human evaluations are conducted to compare performance across models and prompting strategies. Our findings reveal that model architecture and prompt design significantly impact persuasive effectiveness. Among prompting strategies, few-shot prompting consistently produces strong results across models and tasks. While current models exhibit potential for generating persuasive dialogue, their effectiveness varies across domains. Structured prompting thus emerges as a vital factor for enhancing persuasive success in practical applications.
The massive fame of Ethereum as the second largest blockchain has concurrently made it a lucrative target for malicious actors intended to exploit smart contract vulnerabilities for unlawful financial gain. While current smart contract vulnerability detection (SCVD) approaches are effective to a certain limit, they often suffer from limited modality awareness, ineffective to model diverse data distributions, and an inability to model cross-contract semantics. To bridge these gaps, we propose CM2VD, a novel Contrastive Multimodal Multiview Vulnerability Detection framework that integrates contrastive learning and cross-attention-based multimodal feature fusion for robust vulnerability detection. In particular, our framework introduces a novel feature fusion mechanism that jointly captures intra-and inter-modality dependencies, enabling adaptive modeling of vulnerability-specific patterns. Additionally, our model unifies syntactic and semantic representations from diverse modalities at different levels of abstraction to form a comprehensive multiview understanding of contracts. Extensive empirical validation on three real-world smart contract datasets demonstrates that CM2VD consistently outperforms the strongest baseline (CLEAR), achieving up to 2-6% average improvement in F1-Score across multiple vulnerabilities. We further provide an in-depth experimental analysis for the modality-specific contributions, along with robustness, generalizability, and interpretability of the model. Our code is available at our repository 1.
The present study introduces “rock slope instability score (RSIS)” a novel classification system for assessing rock slope stability. It takes into account geological and geotechnical parameters, as well as the impact of human activities and triggering parameters, which have become more frequent due to climate change and few of them have been ignored in existing classifications. The study focuses on rock slopes of various lithologies from the Indian Himalayas. The development of this new classification system is based on the examination of 81 different rock slopes from various states of the Indian Himalayas. Extensive field surveys, rock sampling, geotechnical laboratory tests, and ground measurements have been conducted at the various slope sites to establish a comprehensive scoring system for the stability assessment. The distributions of weightage to each parameter have been considered, corresponding to its degree of impact in causing slope instability. Sensitivity analysis of all defined parameters of RSIS system has revealed that the majority of the parameters exhibit a strong positive correlation, with Pearson correlation coefficients ranging from 0.74 to 0.61. However, two parameters, namely discontinuity dip and the relationship between slope & discontinuity direction, gives moderate relationship with correlation coefficient values of 0.48 and 0.41, respectively. To avoid any designer biasness in the system, several individuals gathered data set at different times. The proposed classification system has demonstrated a strong correlation with the actual slope condition, and it is quite promising. The outcome of RSIS classification for studied 81 slopes classified 2 slopes under stable condition, 21 slopes as partially stable, 44 as unstable, and 14 as completely unstable.
Magmatic intrusion significantly influences the petrophysical and geochemical transformation of coal and is responsible for the alteration in pore geometry, composition, and hydrocarbon generation potential. Our study investigates the impact of igneous intrusion on the petrophysical and organo-geochemical evolution of coal in samples from the Jharia and Raniganj coalfields, Damodar Valley basin. The study follows a multi-pronged approach using low-pressure gas adsorption, Rock-Eval pyrolysis, scanning electron microscopy, maceral analysis and proximate and ultimate analysis. Results show that in proximity to igneous intrusions, coal exhibits intensive thermal alteration with significantly higher thermal maturity (VRr up to 1.406 %) and vitrinite enrichment but a reduced micropore volume, leading to poor gas storage but superior gas seepage. Conversely, unaffected coals with weaker magmatic influence show semi-closed-bottleneck pores and poor connectivity, indicating potential hindrance for gas transport. Approaching the site of intrusion, a 164 per cent increase in vitrinite reflectance is observed, indicating significant thermal alteration. This transformation is accompanied by devolatilization vacuoles, a progressive decline in moisture and volatile matter content, and systematic aromatization of the organic matter. The formation of secondary microfractures and devolatilization vacuoles further modifies pore connectivity, impacting reservoir characteristics and CBM recovery potential. The study highlights the transformative impact of thermal intrusions on gas storage and transport properties of coal, offering critical insights for optimizing CBM extraction and improving mine safety in thermally affected coal seams.
Orbital angular momentum (OAM) in microwave frequencies introduces a promising new dimension for enhancing channel capacity in future 6G wireless communication systems. However, the practical use of OAM waves for long-distance transmission is limited by their inherent beam divergence, which reduces signal strength over distance. This paper proposes a transmissive metasurface lens designed to reduce the divergence of an OAM beam generated by a multimode uniform circular array (UCA) operating at 6 GHz. The design of the metasurface lens is based on the optical converging axicon concept, which is used to achieve the required phase-shift distribution across the metasurface. The UCA generates OAM modes of l = − 1, + 1, and 0 through distinct feed excitations. Comprehensive evaluations comparing the UCA performance with and without the metasurface lens were carried out based on radiation patterns, S11, electric field amplitude, and phase distribution. The results demonstrate a significant improvement: the beam divergence is reduced from 18° to 10°, and the maximum gain is increased from 10.7 dBi to 14.1 dBi. These improvements indicate a more focused and directional OAM beam, enhancing the practicality of OAM-based communication systems for extended-range applications.
