Today's AI systems have human-designed, fixed architectures and cannot autonomously and continuously improve themselves. The advance of AI could itself be automated. If done safely, that would accelerate AI development and allow us to reap its benefits much sooner. Meta-learning can automate the discovery of novel algorithms, but is limited by first-order improvements and the human design of a suitable search space. The Gödel machine proposed a theoretical alternative: a self-improving AI that repeatedly modifies itself in a provably beneficial manner. Unfortunately, proving that most changes are net beneficial is impossible in practice. We introduce the Darwin Gödel Machine (DGM), a self-improving system that iteratively modifies its own code (thereby also improving its ability to modify its own codebase) and empirically validates each change using coding benchmarks. Inspired by Darwinian evolution and open-endedness research, the DGM maintains an archive of generated coding agents. It grows the archive by sampling an agent from it and using a foundation model to create a new, interesting, version of the sampled agent. This open-ended exploration forms a growing tree of diverse, high-quality agents and allows the parallel exploration of many different paths through the search space. Empirically, the DGM automatically improves its coding capabilities (e.g., better code editing tools, long-context window management, peer-review mechanisms), increasing performance on SWE-bench from 20.0% to 50.0%, and on Polyglot from 14.2% to 30.7%. Furthermore, the DGM significantly outperforms baselines without self-improvement or open-ended exploration. All experiments were done with safety precautions (e.g., sandboxing, human oversight). The DGM is a significant step toward self-improving AI, capable of gathering its own stepping stones along paths that unfold into endless innovation.

Multi-Agent Path Finding (MAPF) poses a significant and challenging problem critical for applications in robotics and logistics, particularly due to its combinatorial complexity and the partial observability inherent in realistic environments. Decentralized reinforcement learning methods commonly encounter two substantial difficulties: first, they often yield self-centered behaviors among agents, resulting in frequent collisions, and second, their reliance on complex communication modules leads to prolonged training times, sometimes spanning weeks. To address these challenges, we propose an efficient decentralized planning framework based on the Decision Transformer (DT), uniquely leveraging offline reinforcement learning to substantially reduce training durations from weeks to mere hours. Crucially, our approach effectively handles long-horizon credit assignment and significantly improves performance in scenarios with sparse and delayed rewards. Furthermore, to overcome adaptability limitations inherent in standard RL methods under dynamic environmental changes, we integrate a large language model (GPT-4o) to dynamically guide agent policies. Extensive experiments in both static and dynamically changing environments demonstrate that our DT-based approach, augmented briefly by GPT-4o, significantly enhances adaptability and performance.

E-commerce risk management requires aggregating diverse, deeply embedded web data through multi-step, stateful interactions, which traditional scraping methods and most existing Graphical User Interface (GUI) agents cannot handle. These agents are typically limited to single-step tasks and lack the ability to manage dynamic, interactive content critical for effective risk assessment. To address this challenge, we introduce RISK, a novel framework designed to build and deploy GUI agents for this domain. RISK integrates three components: (1) RISK-Data, a dataset of 8,492 single-step and 2,386 multi-step interaction trajectories, collected through a high-fidelity browser framework and a meticulous data curation process; (2) RISK-Bench, a benchmark with 802 single-step and 320 multi-step trajectories across three difficulty levels for standardized evaluation; and (3) RISK-R1, a R1-style reinforcement fine-tuning framework considering four aspects: (i) Output Format: Updated format reward to enhance output syntactic correctness and task comprehension, (ii) Single-step Level: Stepwise accuracy reward to provide granular feedback during early training stages, (iii) Multi-step Level: Process reweight to emphasize critical later steps in interaction sequences, and (iv) Task Level: Level reweight to focus on tasks of varying difficulty. Experiments show that RISK-R1 outperforms existing baselines, achieving a 6.8% improvement in offline single-step and an 8.8% improvement in offline multi-step. RISK provides a scalable, domain-specific solution for automating complex web interactions, advancing the state of the art in e-commerce risk management. In e-commerce transaction scenarios, stringent compliance and risk control mechanisms are essential to mitigate operational, regulatory, and reputational risks. Decision-making in this context requires the aggregation of heterogeneous information from multiple external sources, many of which exist as unstructured or semi-structured data on the public web. While broad web search can identify relevant sources, truly actionable intelligence often resides deep within specific websites--sometimes on dynamically loaded subpages, behind interactive elements, or embedded within complex document object models (DOM). Traditional scraping APIs or static crawlers fail to retrieve such deeply embedded content, as they lack the ability to engage in stateful, event-driven interactions (Petrova et al., 2025). This work was done when the first author was an intern at Ant International.

Fine-tuning large language models for code editing has typically relied on mining commits and pull requests. The working hypothesis has been that commit messages describe human intent in natural language, and patches to code describe the changes that implement that intent. However, much of the previously collected data is noisy: commit messages are terse, human-written commits commingle several unrelated edits, and many commits come from simple, rule-based bots. The recent adoption of software engineering agents changes this landscape. Code changes co-authored by humans and agents tend to be more narrowly scoped and focused on clearer goals. Their commit messages, generated by LLMs, articulate intent and rationale in much greater detail. Moreover, when these changes land in public repositories, they are implicitly filtered by humans: maintainers discard low-quality commits to their projects. We present AgentPack, a corpus of 1.3M code edits co-authored by Claude Code, OpenAI Codex, and Cursor Agent across public GitHub projects up to mid-August 2025. We describe the identification and curation pipeline, quantify adoption trends of these agents, and analyze the structural properties of the edits. Finally, we show that models fine-tuned on AgentPack can outperform models trained on prior human-only commit corpora, highlighting the potential of using public data from software engineering agents to train future code-editing models.

Conventional maneuver prediction methods use some sort of classification model on temporal trajectory data to predict behavior of agents over a set time horizon. Despite of having the best precision and recall, these models cannot predict a lane change accurately unless they incorporate information about the entire scene. Level-k game theory can leverage the human-like hierarchical reasoning to come up with the most rational decisions each agent can make in a group. This can be leveraged to model interactions between different vehicles in presence of each other and hence compute the most rational decisions each agent would make. The result of game theoretic evaluation can be used as a "prior" or combined with a traditional motion-based classification model to achieve more accurate predictions. The proposed approach assumes that the states of the vehicles around the target lead vehicle are known. The module will output the most rational maneuver prediction of the target vehicle based on an online optimization solution. These predictions are instrumental in decision making systems like Adaptive Cruise Control (ACC) or Traxen's iQ-Cruise further improving the resulting fuel savings.

There is growing interest in using Large Language Models as agents (LLM agents) for social simulations to inform policy, yet real-world adoption remains limited. This paper addresses the question: How can LLM agent simulations be made genuinely useful for policy? We report on a year-long iterative design engagement with a university emergency preparedness team. Across multiple design iterations, we iteratively developed a system of 13,000 LLM agents that simulate crowd movement and communication during a large-scale gathering under various emergency scenarios. These simulations informed actual policy implementation, shaping volunteer training, evacuation protocols, and infrastructure planning. Analyzing this process, we identify three design implications: start with verifiable scenarios and build trust gradually, use preliminary simulations to elicit tacit knowledge, and treat simulation and policy development as evolving together. These implications highlight actionable pathways to making LLM agent simulations that are genuinely useful for policy.

As a model-agnostic approach to long context modeling, multi-agent systems can process inputs longer than a large language model's context window without retraining or architectural modifications. However, their performance often heavily relies on hand-crafted multi-agent collaboration strategies and prompt engineering, which limit generalizability. In this work, we introduce a principled framework that formalizes the model-agnostic long context modeling problem as a compression problem, yielding an information-theoretic compression objective. Building on this framework, we propose Graph of Agents (GoA), which dynamically constructs an input-dependent collaboration structure that maximizes this objective. For Llama 3.1 8B and Qwen3 8B across six document question answering benchmarks, GoA improves the average $F_1$ score of retrieval-augmented generation by 5.7\% and a strong multi-agent baseline using a fixed collaboration structure by 16.35\%, respectively. Even with only a 2K context window, GoA surpasses the 128K context window Llama 3.1 8B on LongBench, showing a dramatic increase in effective context length. Our source code is available at https://github.com/tjoo512/graph-of-agents.

We introduce a framework for modeling decisions with axioms that are statements about decisions, e.g., ethical constraints. Using our framework we define a taxonomy of decision axioms based on their structural properties and demonstrate a tension between the use of axioms to make decisions and the use of axioms to evaluate decisions which we call the Decision-Evaluation Paradox. We argue that the Decision-Evaluation Paradox arises with realistic axiom structures, and the paradox illuminates why one must be exceptionally careful when training models on decision data or applying axioms to make and evaluate decisions.

We study the problem of online multi-agent reinforcement learning (MARL) in environments with sparse rewards, where reward feedback is not provided at each interaction but only revealed at the end of a trajectory. This setting, though realistic, presents a fundamental challenge: the lack of intermediate rewards hinders standard MARL algorithms from effectively guiding policy learning. To address this issue, we propose a novel framework that integrates online inverse preference learning with multi-agent on-policy optimization into a unified architecture. At its core, our approach introduces an implicit multi-agent reward learning model, built upon a preference-based value-decomposition network, which produces both global and local reward signals. These signals are further used to construct dual advantage streams, enabling differentiated learning targets for the centralized critic and decentralized actors. In addition, we demonstrate how large language models (LLMs) can be leveraged to provide preference labels that enhance the quality of the learned reward model. Empirical evaluations on state-of-the-art benchmarks, including MAMuJoCo and SMACv2, show that our method achieves superior performance compared to existing baselines, highlighting its effectiveness in addressing sparse-reward challenges in online MARL.

Developing and testing user interfaces (UIs) and training AI agents to interact with them are challenging due to the dynamic and diverse nature of real-world mobile environments. Existing methods often rely on cumbersome physical devices or limited static analysis of screenshots, which hinders scalable testing and the development of intelligent UI agents. We introduce UISim, a novel image-based UI simulator that offers a dynamic and interactive platform for exploring mobile phone environments purely from screen images. Our system employs a two-stage method: given an initial phone screen image and a user action, it first predicts the abstract layout of the next UI state, then synthesizes a new, visually consistent image based on this predicted layout. This approach enables the realistic simulation of UI transitions. UISim provides immediate practical benefits for UI testing, rapid prototyping, and synthetic data generation. Furthermore, its interactive capabilities pave the way for advanced applications, such as UI navigation task planning for AI agents. Our experimental results show that UISim outperforms end-to-end UI generation baselines in generating realistic and coherent subsequent UI states, highlighting its fidelity and potential to streamline UI development and enhance AI agent training.