Language agents have achieved considerable performance on various complex tasks. Despite the incessant exploration in this field, existing language agent systems still struggle with costly, non-reproducible data reliance and face the challenge of compelling a single model for multiple functions. To this end, we introduce AutoAct, an automatic agent learning framework that does not rely on large-scale annotated data and synthetic trajectories from closed-source models (e.g., GPT-4). Given limited data with a tool library, AutoAct first automatically synthesizes planning trajectories without any assistance from humans or strong closed-source models. Then, AutoAct leverages a division-of-labor strategy to automatically differentiate based on the target task information and synthesized trajectories, producing a sub-agent group to complete the task. We conduct comprehensive experiments with different LLMs, which demonstrates that AutoAct yields better or parallel performance compared to various strong baselines. We even notice that AutoAct, when using the Llama-2-13b model, can achieve performance comparable to that of the zero-shot GPT-3.5-Turbo agent. Code will be available at https://github.com/zjunlp/AutoAct.

Agents centered around Large Language Models (LLMs) are now capable of automating mobile device operations for users. After fine-tuning to learn a user's mobile operations, these agents can adhere to high-level user instructions online. They execute tasks such as goal decomposition, sequencing of sub-goals, and interactive environmental exploration, until the final objective is achieved. However, privacy concerns related to personalized user data arise during mobile operations, requiring user confirmation. Moreover, users' real-world operations are exploratory, with action data being complex and redundant, posing challenges for agent learning. To address these issues, in our practical application, we have designed interactive tasks between agents and humans to identify sensitive information and align with personalized user needs. Additionally, we integrated Standard Operating Procedure (SOP) information within the model's in-context learning to enhance the agent's comprehension of complex task execution. Our approach is evaluated on the new device control benchmark AitW, which encompasses 30K unique instructions across multi-step tasks, including application operation, web searching, and web shopping. Experimental results show that the SOP-based agent achieves state-of-the-art performance in LLMs without incurring additional inference costs, boasting an overall action success rate of 66.92\%. The code and data examples are available at https://github.com/alipay/mobile-agent.

Language is often considered a key aspect of human thinking, providing us with exceptional abilities to generalize, explore, plan, replan, and adapt to new situations. However, Reinforcement Learning (RL) agents are far from human-level performance in any of these abilities. We hypothesize one reason for such cognitive deficiencies is that they lack the benefits of thinking in language and that we can improve AI agents by training them to think like humans do. We introduce a novel Imitation Learning framework, Thought Cloning, where the idea is to not just clone the behaviors of human demonstrators, but also the thoughts humans have as they perform these behaviors. While we expect Thought Cloning to truly shine at scale on internet-sized datasets of humans thinking out loud while acting (e.g. online videos with transcripts), here we conduct experiments in a domain where the thinking and action data are synthetically generated. Results reveal that Thought Cloning learns much faster than Behavioral Cloning and its performance advantage grows the further out of distribution test tasks are, highlighting its ability to better handle novel situations. Thought Cloning also provides important benefits for AI Safety and Interpretability, and makes it easier to debug and improve AI. Because we can observe the agent's thoughts, we can (1) more easily diagnose why things are going wrong, making it easier to fix the problem, (2) steer the agent by correcting its thinking, or (3) prevent it from doing unsafe things it plans to do. Overall, by training agents how to think as well as behave, Thought Cloning creates safer, more powerful agents.

We present a comprehensive, user-centric approach to understand preferences in AI-based productivity agents and develop personalized solutions tailored to users' needs. Utilizing a two-phase method, we first conducted a survey with 363 participants, exploring various aspects of productivity, communication style, agent approach, personality traits, personalization, and privacy. Drawing on the survey insights, we developed a GPT-4 powered personalized productivity agent that utilizes telemetry data gathered via Viva Insights from information workers to provide tailored assistance. We compared its performance with alternative productivity-assistive tools, such as dashboard and narrative, in a study involving 40 participants. Our findings highlight the importance of user-centric design, adaptability, and the balance between personalization and privacy in AI-assisted productivity tools. By building on the insights distilled from our study, we believe that our work can enable and guide future research to further enhance productivity solutions, ultimately leading to optimized efficiency and user experiences for information workers.

Centralized training with decentralized execution (CTDE) is widely employed to stabilize partially observable multi-agent reinforcement learning (MARL) by utilizing a centralized value function during training. However, existing methods typically assume that agents make decisions based on their local observations independently, which may not lead to a correlated joint policy with sufficient coordination. Inspired by the concept of correlated equilibrium, we propose to introduce a \textit{strategy modification} to provide a mechanism for agents to correlate their policies. Specifically, we present a novel framework, AgentMixer, which constructs the joint fully observable policy as a non-linear combination of individual partially observable policies. To enable decentralized execution, one can derive individual policies by imitating the joint policy. Unfortunately, such imitation learning can lead to \textit{asymmetric learning failure} caused by the mismatch between joint policy and individual policy information. To mitigate this issue, we jointly train the joint policy and individual policies and introduce \textit{Individual-Global-Consistency} to guarantee mode consistency between the centralized and decentralized policies. We then theoretically prove that AgentMixer converges to an $\epsilon$-approximate Correlated Equilibrium. The strong experimental performance on three MARL benchmarks demonstrates the effectiveness of our method.

We propose a novel, heterogeneous multi-agent architecture that miniaturizes rovers by outsourcing power generation to a central hub. By delegating power generation and distribution functions to this hub, the size, weight, power, and cost (SWAP-C) per rover are reduced, enabling efficient fleet scaling. As these rovers conduct mission tasks around the terrain, the hub charges an array of replacement battery modules. When a rover requires charging, it returns to the hub to initiate an autonomous docking sequence and exits with a fully charged battery. This confers an advantage over direct charging methods, such as wireless or wired charging, by replenishing a rover in minutes as opposed to hours, increasing net rover uptime. This work shares an open-source platform developed to demonstrate battery swapping on unknown field terrain. We detail our design methodologies utilized for increasing system reliability, with a focus on optimization, robust mechanical design, and verification. Optimization of the system is discussed, including the design of passive guide rails through simulation-based optimization methods which increase the valid docking configuration space by 258%. The full system was evaluated during integrated testing, where an average servicing time of 98 seconds was achieved on surfaces with a gradient up to 10{\deg}. We conclude by briefly proposing flight considerations for advancing the system toward a space-ready design. In sum, this prototype represents a proof of concept for autonomous docking and battery transfer on field terrain, advancing its Technology Readiness Level (TRL) from 1 to 3.

This paper introduces a methodology through which a population of autonomous agents can establish a linguistic convention that enables them to refer to arbitrary entities that they observe in their environment. The linguistic convention emerges in a decentralised manner through local communicative interactions between pairs of agents drawn from the population. The convention consists of symbolic labels (word forms) associated to concept representations (word meanings) that are grounded in a continuous feature space. The concept representations of each agent are individually constructed yet compatible on a communicative level. Through a range of experiments, we show (i) that the methodology enables a population to converge on a communicatively effective, coherent and human-interpretable linguistic convention, (ii) that it is naturally robust against sensor defects in individual agents, (iii) that it can effectively deal with noisy observations, uncalibrated sensors and heteromorphic populations, (iv) that the method is adequate for continual learning, and (v) that the convention self-adapts to changes in the environment and communicative needs of the agents.

This study introduces CycLight, a novel cycle-level deep reinforcement learning (RL) approach for network-level adaptive traffic signal control (NATSC) systems. Unlike most traditional RL-based traffic controllers that focus on step-by-step decision making, CycLight adopts a cycle-level strategy, optimizing cycle length and splits simultaneously using Parameterized Deep Q-Networks (PDQN) algorithm. This cycle-level approach effectively reduces the computational burden associated with frequent data communication, meanwhile enhancing the practicality and safety of real-world applications. A decentralized framework is formulated for multi-agent cooperation, while attention mechanism is integrated to accurately assess the impact of the surroundings on the current intersection. CycLight is tested in a large synthetic traffic grid using the microscopic traffic simulation tool, SUMO. Experimental results not only demonstrate the superiority of CycLight over other state-of-the-art approaches but also showcase its robustness against information transmission delays.

Connected and automated vehicles (CAVs) have emerged as a potential solution to the future challenges of developing safe, efficient, and eco-friendly transportation systems. However, CAV control presents significant challenges, given the complexity of interconnectivity and coordination required among the vehicles. To address this, multi-agent reinforcement learning (MARL), with its notable advancements in addressing complex problems in autonomous driving, robotics, and human-vehicle interaction, has emerged as a promising tool for enhancing the capabilities of CAVs. However, there is a notable absence of current reviews on the state-of-the-art MARL algorithms in the context of CAVs. Therefore, this paper delivers a comprehensive review of the application of MARL techniques within the field of CAV control. The paper begins by introducing MARL, followed by a detailed explanation of its unique advantages in addressing complex mobility and traffic scenarios that involve multiple agents. It then presents a comprehensive survey of MARL applications on the extent of control dimensions for CAVs, covering critical and typical scenarios such as platooning control, lane-changing, and unsignalized intersections. In addition, the paper provides a comprehensive review of the prominent simulation platforms used to create reliable environments for training in MARL. Lastly, the paper examines the current challenges associated with deploying MARL within CAV control and outlines potential solutions that can effectively overcome these issues. Through this review, the study highlights the tremendous potential of MARL to enhance the performance and collaboration of CAV control in terms of safety, travel efficiency, and economy.

This paper addresses motion forecasting in multi-agent environments, pivotal for ensuring safety of autonomous vehicles. Traditional as well as recent data-driven marginal trajectory prediction methods struggle to properly learn non-linear agent-to-agent interactions. We present SSL-Interactions that proposes pretext tasks to enhance interaction modeling for trajectory prediction. We introduce four interaction-aware pretext tasks to encapsulate various aspects of agent interactions: range gap prediction, closest distance prediction, direction of movement prediction, and type of interaction prediction. We further propose an approach to curate interaction-heavy scenarios from datasets. This curated data has two advantages: it provides a stronger learning signal to the interaction model, and facilitates generation of pseudo-labels for interaction-centric pretext tasks. We also propose three new metrics specifically designed to evaluate predictions in interactive scenes. Our empirical evaluations indicate SSL-Interactions outperforms state-of-the-art motion forecasting methods quantitatively with up to 8% improvement, and qualitatively, for interaction-heavy scenarios.