Despite significant progress on multi-agent reinforcement learning (MARL) in recent years, coordination in complex domains remains a challenge. Work in MARL often focuses on solving tasks where agents interact with all other agents and entities in the environment; however, we observe that real-world tasks are often composed of several isolated instances of local agent interactions (subtasks), and each agent can meaningfully focus on one subtask to the exclusion of all else in the environment. In these composite tasks, successful policies can often be decomposed into two levels of decision-making: agents are allocated to specific subtasks and each agent acts productively towards their assigned subtask alone. This decomposed decision making provides a strong structural inductive bias, significantly reduces agent observation spaces, and encourages subtask-specific policies to be reused and composed during training, as opposed to treating each new composition of subtasks as unique. We introduce ALMA, a general learning method for taking advantage of these structured tasks. ALMA simultaneously learns a high-level subtask allocation policy and low-level agent policies. We demonstrate that ALMA learns sophisticated coordination behavior in a number of challenging environments, outperforming strong baselines. ALMA's modularity also enables it to better generalize to new environment configurations. Finally, we find that while ALMA can integrate separately trained allocation and action policies, the best performance is obtained only by training all components jointly.

Multi-agent Large Language Model (LLM) systems have been leading the way in applied LLM research across a number of fields. One notable area is software development, where researchers have advanced the automation of code implementation, code testing, code maintenance, inter alia, using LLM agents. However, software development is a multifaceted environment that extends beyond just code. As such, a successful LLM system must factor in multiple stages of the software development life-cycle (SDLC). In this paper, we propose a vision for ALMAS, an Autonomous LLM-based Multi-Agent Software Engineering framework, which follows the above SDLC philosophy such that it may work within an agile software development team to perform several tasks end-to-end. ALMAS aligns its agents with agile roles, and can be used in a modular fashion to seamlessly integrate with human developers and their development environment. We showcase the progress towards ALMAS through our published works and a use case demonstrating the framework, where ALMAS is able to seamlessly generate an application and add a new feature.

Can Israel survive economic isolation? The activists of the Global Sumud Flotilla remain alert. Wednesday's drone attack on the vessels - heading towards Gaza to break Israel's siege on the Palestinian enclave - is not expected to be the last. As the flotilla, currently travelling in Greek territorial waters, nears Gaza, a larger Israeli attack is expected. Acar is experienced in spotting drones and assessing security risks, having previous experience trying to reach Gaza on June's Madleen flotilla, which Israel intercepted.

The increasing volume of papers and proposals undergoing peer review emphasizes the pressing need for greater automation to effectively manage the growing scale. In this study, we present the deployment and evaluation of machine learning and optimization techniques for assigning proposals to reviewers that was developed for the Atacama Large Millimeter/submillimeter Array (ALMA) during the Cycle 10 Call for Proposals issued in 2023. By utilizing topic modeling algorithms, we identify the proposal topics and assess reviewers' expertise based on their historical ALMA proposal submissions. We then apply an adapted version of the assignment optimization algorithm from PeerReview4All (Stelmakh et al. 2021a) to maximize the alignment between proposal topics and reviewer expertise. Our evaluation shows a significant improvement in matching reviewer expertise: the median similarity score between the proposal topic and reviewer expertise increased by 51 percentage points compared to the previous cycle, and the percentage of reviewers reporting expertise in their assigned proposals rose by 20 percentage points. Furthermore, the assignment process proved highly effective in that no proposals required reassignment due to significant mismatches, resulting in a savings of 3 to 5 days of manual effort.

Despite significant progress on multi-agent reinforcement learning (MARL) in recent years, coordination in complex domains remains a challenge. Work in MARL often focuses on solving tasks where agents interact with all other agents and entities in the environment; however, we observe that real-world tasks are often composed of several isolated instances of local agent interactions (subtasks), and each agent can meaningfully focus on one subtask to the exclusion of all else in the environment. In these composite tasks, successful policies can often be decomposed into two levels of decision-making: agents are allocated to specific subtasks and each agent acts productively towards their assigned subtask alone. This decomposed decision making provides a strong structural inductive bias, significantly reduces agent observation spaces, and encourages subtask-specific policies to be reused and composed during training, as opposed to treating each new composition of subtasks as unique. We introduce ALMA, a general learning method for taking advantage of these structured tasks.

Despite significant progress on multi-agent reinforcement learning (MARL) in recent years, coordination in complex domains remains a challenge. Work in MARL often focuses on solving tasks where agents interact with all other agents and entities in the environment; however, we observe that real-world tasks are often composed of several isolated instances of local agent interactions (subtasks), and each agent can meaningfully focus on one subtask to the exclusion of all else in the environment. In these composite tasks, successful policies can often be decomposed into two levels of decision-making: agents are allocated to specific subtasks and each agent acts productively towards their assigned subtask alone. This decomposed decision making provides a strong structural inductive bias, significantly reduces agent observation spaces, and encourages subtask-specific policies to be reused and composed during training, as opposed to treating each new composition of subtasks as unique. We introduce ALMA, a general learning method for taking advantage of these structured tasks. ALMA simultaneously learns a high-level subtask allocation policy and low-level agent policies. We demonstrate that ALMA learns sophisticated coordination behavior in a number of challenging environments, outperforming strong baselines. ALMA's modularity also enables it to better generalize to new environment configurations. Finally, we find that while ALMA can integrate separately trained allocation and action policies, the best performance is obtained only by training all components jointly.

In many practical applications of machine learning data arrives sequentially over time in large chunks. Practitioners have then to decide how to allocate their computational budget in order to obtain the best performance at any point in time. Online learning theory for convex optimization suggests that the best strategy is to use data as soon as it arrives. However, this might not be the best strategy when using deep non-linear networks, particularly when these perform multiple passes over each chunk of data rendering the overall distribution non i.i.d.. In this paper, we formalize this learning setting in the simplest scenario in which each data chunk is drawn from the same underlying distribution, and make a first attempt at empirically answering the following questions: How long should the learner wait before training on the newly arrived chunks? What architecture should the learner adopt? Should the learner increase capacity over time as more data is observed? We probe this learning setting using convolutional neural networks trained on classic computer vision benchmarks as well as a large transformer model trained on a large-scale language modeling task. Code is available at \url{www.github.com/facebookresearch/ALMA}.

'I'm Your Man' is an unconventional portrait of a woman who empirically lives with a humanoid robot and finds the answer to what makes someone human is not as black-and-white as she once thought. As the world evolves, technology continues to become further ingrained in every aspect of our lives. From the everyday of using a touchscreen cash register to the more advanced instruments used for microsurgery, these tools improve speed, convenience and accuracy. However, the integration into our personal lives has progressed slower for many, keeping it at arm's length for various reasons. Yet, countless people have turned to online dating websites and algorithms to find their perfect match, but for some finding the one seems like an endless journey of energy consumption and heartbreak.

Maria Schrader’s film, starring Dan Stevens as a robot designed to be the perfect man, confirms comedy as the playground of philosophy: nothing is funnier or more stirring than the sight of somebody learning how to be.

Directed by Maria Schrader, this was a crowd-pleasing favourite at the Berlin film festival earlier this year and its star, Maren Eggert, won the festival's new gender-neutral best leading performance prize. But I was disappointed with a film whose crises and dilemmas seem laborious and essentially predictable; it does not fully work as sci-fi or satire or comedy. We are in a world of the near-future (and the city of Berlin itself is certainly very plausible as its location). Eggert plays Alma, an archaeologist with an unhappy and frustrating personal life. She is persuaded by her boss to be a guinea-pig for a new hi-tech scheme: she will road-test a male "companion" robot, programmed to be infinitely considerate and obliging, which will attend to all her emotional and indeed physical needs.