The convergence of embodied intelligence and world models has catalyzed growing interest in integrating physical laws into AI systems. While prior surveys have examined world models and embodied intelligence separately, we focus on the progression that connects these capabilities as a unified developmental pathway from passive observation to active physical comprehension. This survey provides a systematic framework revealing how physical AI advances through four interconnected stages: perception transforms sensory data into structured physical representations, reasoning derives explanations from observed phenomena, modeling enables predictive simulation grounded in physical principles, and embodied interaction closes the loop through physical manipulation and environmental feedback. Each stage enables and enhances the next: perceptual grounding supports causal reasoning, reasoning unlocks predictive capabilities, and robust models drive genuine physical interaction. Through analysis of developments spanning architectural innovations, training methodologies, causal inference, and embodied systems, we synthesize how physical understanding emerges through cumulative integration across this progression. Our framework reveals the evolution from isolated, task-specific solutions toward integrated architectures that advance from pattern recognition toward causal reasoning and counterfactual prediction. This perspective provides foundations for next-generation physical AI systems with direct implications for safe, generalizable, and interpretable deployment across robotics, scientific discovery, and autonomous systems. We maintain a continuously updated taxonomy repository at https://github.com/AI4Phys/Awesome-AI-for-Physics.

In this paper, we address the challenging task of multimodal reasoning by incorporating the notion of “slow thinking” into multimodal large language models (MLLMs). Our core idea is that models can learn to adaptively use different levels of reasoning to tackle questions of varying complexity. We propose a novel paradigm of Self-structured Chain of Thought (SCoT), which consists of minimal semantic atomic steps. Unlike existing methods that rely on structured templates or free-form paradigms, our method not only generates flexible CoT structures for various complex tasks but also mitigates the phenomenon of overthinking for easier tasks. To introduce structured reasoning into visual cognition, we design a novel AtomThink framework with four key modules: (i) a data engine to generate high-quality multimodal reasoning paths; (ii) a supervised fine-tuning (SFT) process with serialized inference data; (iii) a policy-guided multi-turn inference method; and (iv) an atomic capability metric to evaluate the single-step utilization rate. Extensive experiments demonstrate that the proposed AtomThink significantly improves the performance of baseline MLLMs, achieving more than 10% average accuracy gains on MathVista and MathVerse. Compared to state-of-the-art structured CoT approaches, our method not only achieves higher accuracy but also improves data utilization by 5 × and boosts inference efficiency by 85.3%.

Modern scientific progress is increasingly driven by collaborative endeavors that leverage specialized expertise and constructive peer critique. Multi-agent systems (MAS) offer a robust framework to emulate these collaborative dynamics inherent to human researcher teams by combining distributed information processing with discussion-driven validation, enabling collective intelligence that exceeds the capabilities of individual agents in addressing complex interdisciplinary challenges. We introduce an application-oriented taxonomy that maps canonical stages of the research workflow to both the promise and the current reality of MAS in scientific discovery, providing a coherent foundation for understanding, evaluating, and advancing autonomous AI co-scientists. We highlight the distinctive advantages of MAS over single-agent approaches, identify key bottlenecks limiting current deployments, and outline critical research frontiers to bridge the gap between potential and practice. We argue that MAS hold transformative promise to move beyond the role of assistive tools, evolving into autonomous co-scientists capable of parallel exploration of vast knowledge spaces and robust validation through diverse perspectives, thereby advancing open-ended scientific research in partnership alongside human investigators.