Recently, quadrupedal locomotion has achieved significant success, but their manipulation capabilities, particularly in handling large objects, remain limited, restricting their usefulness in demanding real-world applications such as search and rescue, construction, industrial automation, and room organization. This paper tackles the task of obstacle-aware, long-horizon pushing by multiple quadrupedal robots. We propose a hierarchical multi-agent reinforcement learning framework with three levels of control. The high-level controller integrates an RRT planner and a centralized adaptive policy to generate subgoals, while the mid-level controller uses a decentralized goal-conditioned policy to guide the robots toward these sub-goals. A pre-trained low-level locomotion policy executes the movement commands. We evaluate our method against several baselines in simulation, demonstrating significant improvements over baseline approaches, with 36.0% higher success rates and 24.5% reduction in completion time than the best baseline. Our framework successfully enables long-horizon, obstacle-aware manipulation tasks like Push-Cuboid and Push-T on Go1 robots in the real world.

Vision-language models (VLMs) have excelled in multimodal tasks, but adapting them to embodied decision-making in open-world environments presents challenges. One critical issue is bridging the gap between discrete entities in low-level observations and the abstract concepts required for effective planning. A common solution is building hierarchical agents, where VLMs serve as high-level reasoners that break down tasks into executable sub-tasks, typically specified using language. However, language suffers from the inability to communicate detailed spatial information. We propose visual-temporal context prompting, a novel communication protocol between VLMs and policy models. This protocol leverages object segmentation from past observations to guide policy-environment interactions. Using this approach, we train ROCKET-1, a low-level policy that predicts actions based on concatenated visual observations and segmentation masks, supported by real-time object tracking from SAM-2. Our method unlocks the potential of VLMs, enabling them to tackle complex tasks that demand spatial reasoning. Experiments in Minecraft show that our approach enables agents to achieve previously unattainable tasks, with a $\mathbf{76}\%$ absolute improvement in open-world interaction performance. Codes and demos are now available on the project page: https://craftjarvis.github.io/ROCKET-1.

Agentic AI is not just another tool in the scientific toolkit but a paradigm shift: by allowing autonomous systems to not only collect and process data but also to independently hypothesize, experiment, and even make decisions, agentic AI could fundamentally change how we approach biology. To understand why agentic AI holds so much promise, we first need to grapple with the scale of the challenge. Biological systems, particularly human ones, are incredibly complex--layered, dynamic, and interdependent. Take the immune system, for example. It simultaneously operates across multiple levels, from individual molecules to entire organs, adapting and responding to internal and external stimuli in real-time.

One of the challenges artificial intelligence (AI) faces is how a collection of agents coordinate their behaviour to achieve goals that are not reachable by any single agent. In a recent article by Ozmen et al this was framed as one of six grand challenges: That AI needs to respect human cognitive processes at the human-AI interaction frontier. We suggest that this extends to the AI-AI frontier and that it should also reflect human psychology, as it is the only successful framework we have from which to build out. In this extended abstract we first make the case for collective intelligence in a general setting, drawing on recent work from single neuron complexity in neural networks and ant network adaptability in ant colonies. From there we introduce how species relate to one another in an ecological network via niche selection, niche choice, and niche conformity with the aim of forming an analogy with human social network development as new agents join together and coordinate. From there we show how our social structures are influenced by our neuro-physiology, our psychology, and our language. This emphasises how individual people within a social network influence the structure and performance of that network in complex tasks, and that cognitive faculties such as Theory of Mind play a central role. We finish by discussing the current state of the art in AI and where there is potential for further development of a socially embodied collective artificial intelligence that is capable of guiding its own social structures.

Collective Intelligence plays a central role in a large variety of fields, from economics and evolutionary theory to neural networks and eusocial insects, and it is also core to much of the work on emergence and self-organisation in complex systems theory. However, in human collective intelligence there is still much more to be understood in the relationship between specific psychological processes at the individual level and the emergence of self-organised structures at the social level. Previously psychological factors have played a relatively minor role in the study of collective intelligence as the principles are often quite general and applicable to humans just as readily as insects or other agents without sophisticated psychologies. In this article we emphasise, with examples from other complex adaptive systems, the broad applicability of collective intelligence principles while the mechanisms and time-scales differ significantly between examples. We contend that flexible collective intelligence in human social settings is improved by our use of a specific cognitive tool: our Theory of Mind. We identify several key characteristics of psychologically mediated collective intelligence and show that the development of a Theory of Mind is a crucial factor distinguishing social collective intelligence from general collective intelligence. We then place these capabilities in the context of the next steps in artificial intelligence embedded in a future that includes an effective human-AI hybrid social ecology.

Innate values describe agents' intrinsic motivations, which reflect their inherent interests and preferences to pursue goals and drive them to develop diverse skills satisfying their various needs. The essence of reinforcement learning (RL) is learning from interaction based on reward-driven behaviors, much like natural agents. It is an excellent model to describe the innate-values-driven (IV) behaviors of AI agents. Especially developing the awareness of the AI agent through balancing internal and external utilities based on its needs in different tasks is a crucial problem for individuals learning to support AI agents integrating human society with safety and harmony in the long term. This paper proposes a hierarchical compound intrinsic value reinforcement learning model -- innate-values-driven reinforcement learning termed IVRL to describe the complex behaviors of AI agents' interaction. We formulated the IVRL model and proposed two IVRL models: DQN and A2C. By comparing them with benchmark algorithms such as DQN, DDQN, A2C, and PPO in the Role-Playing Game (RPG) reinforcement learning test platform VIZDoom, we demonstrated that rationally organizing various individual needs can effectively achieve better performance.

We propose a learning-based Control Barrier Function (CBF) to reduce conservatism in collision avoidance of car-like robots. Traditional CBFs often use Euclidean distance between robots' centers as safety margin, neglecting headings and simplifying geometries to circles. While this ensures smooth, differentiable safety functions required by CBFs, it can be overly conservative in tight environments. To address this limitation, we design a heading-aware safety margin that accounts for the robots' orientations, enabling a less conservative and more accurate estimation of safe regions. Since the function computing this safety margin is non-differentiable, we approximate it with a neural network to ensure differentiability and facilitate integration with CBFs. We describe how we achieve bounded learning error and incorporate the upper bound into the CBF to provide formal safety guarantees through forward invariance. We show that our CBF is a high-order CBF with relative degree two for a system with two robots whose dynamics are modeled by the nonlinear kinematic bicycle model. Experimental results in overtaking and bypassing scenarios reveal a 33.5 % reduction in conservatism compared to traditional methods, while maintaining safety. Code: https://github.com/bassamlab/sigmarl

As financial markets grow increasingly complex, there is a rising need for automated tools that can effectively assist human analysts in equity research, particularly within sell-side research. While Generative AI (GenAI) has attracted significant attention in this field, existing AI solutions often fall short due to their narrow focus on technical factors and limited capacity for discretionary judgment. These limitations hinder their ability to adapt to new data in real-time and accurately assess risks, which diminishes their practical value for investors. This paper presents FinRobot, the first AI agent framework specifically designed for equity research. FinRobot employs a multi-agent Chain of Thought (CoT) system, integrating both quantitative and qualitative analyses to emulate the comprehensive reasoning of a human analyst. The system is structured around three specialized agents: the Data-CoT Agent, which aggregates diverse data sources for robust financial integration; the Concept-CoT Agent, which mimics an analysts reasoning to generate actionable insights; and the Thesis-CoT Agent, which synthesizes these insights into a coherent investment thesis and report. FinRobot provides thorough company analysis supported by precise numerical data, industry-appropriate valuation metrics, and realistic risk assessments. Its dynamically updatable data pipeline ensures that research remains timely and relevant, adapting seamlessly to new financial information. Unlike existing automated research tools, such as CapitalCube and Wright Reports, FinRobot delivers insights comparable to those produced by major brokerage firms and fundamental research vendors. We open-source FinRobot at \url{https://github. com/AI4Finance-Foundation/FinRobot}.

Differential equations offer a foundational yet powerful framework for modeling interactions within complex dynamic systems and are widely applied across numerous scientific fields. One common challenge in this area is estimating the unknown parameters of these dynamic relationships. However, traditional numerical optimization methods rely on the selection of initial parameter values, making them prone to local optima. Meanwhile, deep learning and Bayesian methods require training models on specific differential equations, resulting in poor versatility. This paper reformulates the parameter estimation problem of differential equations as an optimization problem by introducing the concept of particles from the particle swarm optimization algorithm. Building on reinforcement learning-based particle swarm optimization (RLLPSO), this paper proposes a novel method, DERLPSO, for estimating unknown parameters of differential equations. We compared its performance on three typical ordinary differential equations with the state-of-the-art methods, including the RLLPSO algorithm, traditional numerical methods, deep learning approaches, and Bayesian methods. The experimental results demonstrate that our DERLPSO consistently outperforms other methods in terms of performance, achieving an average Mean Square Error of 1.13e-05, which reduces the error by approximately 4 orders of magnitude compared to other methods. Apart from ordinary differential equations, our DERLPSO also show great promise for estimating unknown parameters of partial differential equations. The DERLPSO method proposed in this paper has high accuracy, is independent of initial parameter values, and possesses strong versatility and stability. This work provides new insights into unknown parameter estimation for differential equations.

Vision-and-Language Navigation (VLN), as a widely discussed research direction in embodied intelligence, aims to enable embodied agents to navigate in complicated visual environments through natural language commands. Most existing VLN methods focus on indoor ground robot scenarios. However, when applied to UAV VLN in outdoor urban scenes, it faces two significant challenges. First, urban scenes contain numerous objects, which makes it challenging to match fine-grained landmarks in images with complex textual descriptions of these landmarks. Second, overall environmental information encompasses multiple modal dimensions, and the diversity of representations significantly increases the complexity of the encoding process. To address these challenges, we propose NavAgent, the first urban UAV embodied navigation model driven by a large Vision-Language Model. NavAgent undertakes navigation tasks by synthesizing multi-scale environmental information, including topological maps (global), panoramas (medium), and fine-grained landmarks (local). Specifically, we utilize GLIP to build a visual recognizer for landmark capable of identifying and linguisticizing fine-grained landmarks. Subsequently, we develop dynamically growing scene topology map that integrate environmental information and employ Graph Convolutional Networks to encode global environmental data. In addition, to train the visual recognizer for landmark, we develop NavAgent-Landmark2K, the first fine-grained landmark dataset for real urban street scenes. In experiments conducted on the Touchdown and Map2seq datasets, NavAgent outperforms strong baseline models. The code and dataset will be released to the community to facilitate the exploration and development of outdoor VLN.