The main challenge of multiagent reinforcement learning is the difficulty of learning useful policies in the presence of other simultaneously learning agents whose changing behaviors jointly affect the environment's transition and reward dynamics. An effective approach that has recently emerged for addressing this non-stationarity is for each agent to anticipate the learning of other agents and influence the evolution of future policies towards desirable behavior for its own benefit. Unfortunately, previous approaches for achieving this suffer from myopic evaluation, considering only a finite number of policy updates. As such, these methods can only influence transient future policies rather than achieving the promise of scalable equilibrium selection approaches that influence the behavior at convergence. In this paper, we propose a principled framework for considering the limiting policies of other agents as time approaches infinity.

Millions of embryos created through IVF sit frozen in time, stored in cryopreservation tanks around the world. The number is only growing thanks to advances in technology, the rising popularity of IVF, and improvements in its success rates. At a basic level, an embryo is simply a tiny ball of a hundred or so cells. But unlike other types of body tissue, it holds the potential for life. Many argue that this endows embryos with a special moral status, one that requires special protections.

As AI continues to advance, there is a growing demand for systems that go beyond language-based assistance and move toward intelligent agents capable of performing real-world actions. This evolution requires the transition from traditional Large Language Models (LLMs), which excel at generating textual responses, to Large Action Models (LAMs), designed for action generation and execution within dynamic environments. Enabled by agent systems, LAMs hold the potential to transform AI from passive language understanding to active task completion, marking a significant milestone in the progression toward artificial general intelligence. In this paper, we present a comprehensive framework for developing LAMs, offering a systematic approach to their creation, from inception to deployment. We begin with an overview of LAMs, highlighting their unique characteristics and delineating their differences from LLMs. Using a Windows OS-based agent as a case study, we provide a detailed, step-by-step guide on the key stages of LAM development, including data collection, model training, environment integration, grounding, and evaluation. This generalizable workflow can serve as a blueprint for creating functional LAMs in various application domains. We conclude by identifying the current limitations of LAMs and discussing directions for future research and industrial deployment, emphasizing the challenges and opportunities that lie ahead in realizing the full potential of LAMs in real-world applications. The code for the data collection process utilized in this paper is publicly available at: https://github.com/microsoft/UFO/tree/main/dataflow, and comprehensive documentation can be found at https://microsoft.github.io/UFO/dataflow/overview/.

Multi-agent frameworks powered by large language models (LLMs) have demonstrated great success in automated planning and task execution. However, the effective adjustment of Agentic workflows during execution has not been well-studied. A effective workflow adjustment is crucial, as in many real-world scenarios, the initial plan must adjust to unforeseen challenges and changing conditions in real-time to ensure the efficient execution of complex tasks. In this paper, we define workflows as an activity-on-vertex (AOV) graphs. We continuously refine the workflow by dynamically adjusting task allocations based on historical performance and previous AOV with LLM agents. To further enhance system performance, we emphasize modularity in workflow design based on measuring parallelism and dependence complexity. Our proposed multi-agent framework achieved efficient sub-task concurrent execution, goal achievement, and error tolerance. Empirical results across different practical tasks demonstrate dramatic improvements in the efficiency of multi-agent frameworks through dynamic workflow updating and modularization.

Recent advances in prompting techniques and multi-agent systems for Large Language Models (LLMs) have produced increasingly complex approaches. However, we lack a framework for characterizing and comparing prompting techniques or understanding their relationship to multi-agent LLM systems. This position paper introduces and explains the concepts of linear contexts (a single, continuous sequence of interactions) and non-linear contexts (branching or multi-path) in LLM systems. These concepts enable the development of an agent-centric projection of prompting techniques, a framework that can reveal deep connections between prompting strategies and multi-agent systems. We propose three conjectures based on this framework: (1) results from non-linear prompting techniques can predict outcomes in equivalent multi-agent systems, (2) multi-agent system architectures can be replicated through single-LLM prompting techniques that simulate equivalent interaction patterns, and (3) these equivalences suggest novel approaches for generating synthetic training data. We argue that this perspective enables systematic cross-pollination of research findings between prompting and multi-agent domains, while providing new directions for improving both the design and training of future LLM systems.

Leveraging large language models (LLMs), an agent can utilize retrieval-augmented generation (RAG) techniques to integrate external knowledge and increase the reliability of its responses. Current RAG-based agents integrate single, domain-specific knowledge sources, limiting their ability and leading to hallucinated or inaccurate responses when addressing cross-domain queries. Integrating multiple knowledge bases into a unified RAG-based agent raises significant challenges, including increased retrieval overhead and data sovereignty when sensitive data is involved. In this work, we propose RopMura, a novel multi-agent system that addresses these limitations by incorporating highly efficient routing and planning mechanisms. RopMura features two key components: a router that intelligently selects the most relevant agents based on knowledge boundaries and a planner that decomposes complex multi-hop queries into manageable steps, allowing for coordinating cross-domain responses. Experimental results demonstrate that RopMura effectively handles both single-hop and multi-hop queries, with the routing mechanism enabling precise answers for single-hop queries and the combined routing and planning mechanisms achieving accurate, multi-step resolutions for complex queries.

Since the emergence of the LLM trend, initiated with the first release of ChatGPT [1], these systems have undergone continuous development and have evolved into multimodal architectures. Multimodal models, such as GPT-4o [2], LLaMA 3.2 [3] and Claude with its latest 3.5 Sonnet model [4], integrate language understanding with non-language capabilities, including vision and audio processing. This progression unlocks new opportunities for developing intelligent agents capable of recognizing and interpreting patterns not only at a semantic level but also through components that can incorporate other types of unstructured data into prompts, significantly expanding their potential applications and impact. Extending these capabilities, Vision-Language Models (VLMs) build on multimodal principles by integrating visual reasoning into the LLM framework. By introducing new tokens into the prompts to process image frames, VLMs enable simultaneous semantic and visual reasoning. This enhancement is particularly valuable in dynamic applications like robotics, where the integration of vision and language reasoning enables systems to generate environment-responsive actions. Such actions, often described as descriptive policies, translate reasoning into meaningful, executable commands. Language models able to generate such commands are usually referred to as "agentic".

In recent years, researchers introduced the Multi-Agent Path Finding in Continuous Time (MAPFR) problem. Conflict-based search with Continuous Time (CCBS), a variant of CBS for discrete MAPF, aims to solve MAPFR with completeness and optimality guarantees. However, CCBS overlooked the fact that search algorithms only guarantee termination and return the optimal solution with a finite amount of search nodes. In this paper, we show that CCBS is incomplete, reveal the gaps in the existing implementation, demonstrate that patching is non-trivial, and discuss the next steps.

Agent-based program repair offers to automatically resolve complex bugs end-to-end by combining the planning, tool use, and code generation abilities of modern LLMs. Recent work has explored the use of agent-based repair approaches on the popular open-source SWE-Bench, a collection of bugs from highly-rated GitHub Python projects. In addition, various agentic approaches such as SWE-Agent have been proposed to solve bugs in this benchmark. This paper explores the viability of using an agentic approach to address bugs in an enterprise context. To investigate this, we curate an evaluation set of 178 bugs drawn from Google's issue tracking system. This dataset spans both human-reported (78) and machine-reported bugs (100). To establish a repair performance baseline on this benchmark, we implement Passerine, an agent similar in spirit to SWE-Agent that can work within Google's development environment. We show that with 20 trajectory samples and Gemini 1.5 Pro, Passerine can produce a patch that passes bug tests (i.e., plausible) for 73% of machine-reported and 25.6% of human-reported bugs in our evaluation set. After manual examination, we found that 43% of machine-reported bugs and 17.9% of human-reported bugs have at least one patch that is semantically equivalent to the ground-truth patch. These results establish a baseline on an industrially relevant benchmark, which as we show, contains bugs drawn from a different distribution -- in terms of language diversity, size, and spread of changes, etc. -- compared to those in the popular SWE-Bench dataset.

The increasing level of autonomy of robots poses challenges of trust and social acceptance, especially in human-robot interaction scenarios. This requires an interpretable implementation of robotic cognitive capabilities, possibly based on formal methods as logics for the definition of task specifications. However, prior knowledge is often unavailable in complex realistic scenarios. In this paper, we propose an offline algorithm based on inductive logic programming from noisy examples to extract task specifications (i.e., action preconditions, constraints and effects) directly from raw data of few heterogeneous (i.e., not repetitive) robotic executions. Our algorithm leverages on the output of any unsupervised action identification algorithm from video-kinematic recordings. Combining it with the definition of very basic, almost task-agnostic, commonsense concepts about the environment, which contribute to the interpretability of our methodology, we are able to learn logical axioms encoding preconditions of actions, as well as their effects in the event calculus paradigm. Since the quality of learned specifications depends mainly on the accuracy of the action identification algorithm, we also propose an online framework for incremental refinement of task knowledge from user feedback, guaranteeing safe execution. Results in a standard manipulation task and benchmark for user training in the safety-critical surgical robotic scenario, show the robustness, data- and time-efficiency of our methodology, with promising results towards the scalability in more complex domains.