Understanding continuance intention of social Q&A communities for informal learning among university students

Abstract

Social Q&A refers to the process of information seeking based on questioning‐and‐answering in natural language through social networks. It facilitates knowledge gathering and interacting, and thus can be regarded as an appropriate environment for informal learning. This study intends to explore factors affecting continuance intention of using social Q&A communities for informal learning and meanwhile reveal the underlying mechanism among these factors. Data were obtained from 257 university students through an online survey. By using partial least squares based structural equation model, the proposed hypotheses and research model were analyzed. The findings show that homophily, interactivity and social presence are three determinants of collaborative learning and community identification in social Q&A community which further influences university students' continuance intention for informal learning. Our research extends the understanding of informal learning in social Q&A communities and indicates how to promote continuous informal learning in such environment.

A Longchain Approach to Reduce Hallucinations in Large Language Models

“…The implementation of pre-training on large, diverse datasets helped LLMs develop a broader knowledge base, which contributed to reducing inaccuracies in their outputs [16]. Transformer-based architectures, with their ability to capture long-range dependencies, were instrumental in enhancing the coherence and factual correctness of generated text [17], [18]. The use of human feedback loops, where model outputs were iteratively refined based on human input, significantly improved the quality and accuracy of responses [19].…”

Evaluating Abstract Reasoning and Problem-Solving Abilities of Large Language Models Using Raven's Progressive Matrices

“…Adaptations of traditional IQ tests to digital formats facilitated the assessment of AI reasoning in a structured manner [48]- [50]. Interactive reasoning tasks that required AI to engage in simulated dialogues demonstrated their capacity for dynamic problem-solving [22], [51]. Assessments incorporating elements of spatial reasoning provided insights into the models' ability to understand and manipulate spatial relationships [52], [53].…”

Optimizing Large Language Model Scaling with Micro Batch Pipeline and Inference Parallelism

“…Mixed precision training further augmented the scalability of LLMs through the utilization of lower precision arithmetic to accelerate computations and reduce memory consumption without sacrificing model accuracy [11,12,13]. Mixed precision training achieved substantial speed-ups and allowed the training of larger models within the same hardware constraints, thereby enhancing the overall efficiency of the training process [14,15].…”

“…Micro-batching emerged as a complementary technique to inference parallelism, allowing smaller batches of data to be processed in parallel across different devices [26]. This approach improved hardware utilization and throughput through the minimization of idle times and the maximization of the parallel processing capabilities of the hardware [12,27,28]. The adoption of micro-batching led to substantial improvements in inference efficiency, particularly in scenarios with high concurrency demands [29].…”