Large language models (LLMs) have become fundamental tools for tasks such as question-answering (QA) and text summarization. These models excel at processing long and complex texts, with capacities reaching over 100,000 tokens. As LLMs are popular for handling large-context tasks, ensuring their reliability and accuracy becomes more pressing. Users rely on LLMs to sift through… →

Graph Neural Networks (GNNs) have emerged as the leading approach for graph learning tasks across various domains, including recommender systems, social networks, and bioinformatics. However, GNNs have shown vulnerability to adversarial attacks, particularly structural attacks that modify graph edges. These attacks pose significant challenges in scenarios where attackers have limited access to entity relationships. Despite… →

The paper “MemLong: Memory-Augmented Retrieval for Long Text Modeling” addresses a critical limitation regarding the ability to process long contexts in the field of Large Language Models (LLMs). While LLMs have shown remarkable success in various applications, they struggle with long-sequence tasks due to traditional attention mechanisms’ quadratic time and space complexity. The increasing memory… →

The dynamics governing multi-agent systems (MAS) are complicated and frequently unknown, making the identification of their underlying graph structures a considerable difficulty. Numerous real-world applications, from robotic swarms to distributed sensor networks, use multi-agent systems, which are made up of autonomous agents interacting in a network. Comprehending the network architecture of these systems is essential… →

The primary challenge in scaling large-scale AI systems is achieving efficient decision-making while maintaining performance. Distributed AI, particularly multi-agent reinforcement learning (MARL), offers potential by decomposing complex tasks and distributing them across collaborative nodes. However, real-world applications face limitations due to high communication and data requirements. Traditional methods, like model predictive control (MPC), require precise… →

Hallucination is a phenomenon where large language models (LLMs) produce responses that are not grounded in reality or do not align with the provided context, generating incorrect, misleading, or nonsensical information. These errors can have serious consequences, particularly in applications that require high precision, like medical diagnosis, legal advice, or other high-stakes scenarios. As the… →

DeepSeek-AI has released DeepSeek-V2.5, a powerful Mixture of Experts (MOE) model with 238 billion parameters, featuring 160 experts and 16 billion active parameters for optimized performance. The model excels in chat and coding tasks, with cutting-edge capabilities such as function calls, JSON output generation, and Fill-in-the-Middle (FIM) completion. With an impressive 128k context length, DeepSeek-V2.5… →

Integrating advanced predictive models into autonomous driving systems has become crucial for enhancing safety and efficiency. Camera-based video prediction emerges as a pivotal component, offering rich real-world data. Content generated by artificial intelligence is presently a leading area of study within the domains of computer vision and artificial intelligence. However, generating photo-realistic and coherent videos… →

Document conversion, particularly from PDF to machine-processable formats, has long presented significant challenges due to PDF files’ diverse and often complex nature. These documents, widely used across various industries, frequently need more standardization, resulting in a loss of structural features when optimized for printing. This structural loss complicates the recovery process, as important elements such… →

Machine learning models, especially those designed for code generation, heavily depend on high-quality data during pretraining. This field has seen rapid advancement, with large language models (LLMs) trained on extensive datasets containing code from various sources. The challenge for researchers is to ensure that the data used is abundant and of high quality, as this… →