While AI systems demonstrate exponentially improving capabilities, the pace of AI research itself remains linearly bounded by human cognitive capacity, creating an increasingly severe development bottleneck. We present ASI-Arch, the first demonstration of Artificial Superintelligence for AI research (ASI4AI) in the critical domain of neural architecture discovery--a fully autonomous system that shatters this fundamental constraint by enabling AI to conduct its own architectural innovation. Moving beyond traditional Neural Architecture Search (NAS), which is fundamentally limited to exploring human-defined spaces, we introduce a paradigm shift from automated optimization to automated innovation. ASI-Arch can conduct end-to-end scientific research in the domain of architecture discovery, autonomously hypothesizing novel architectural concepts, implementing them as executable code, training and empirically validating their performance through rigorous experimentation and past experience. ASI-Arch conducted 1,773 autonomous experiments over 20,000 GPU hours, culminating in the discovery of 106 innovative, state-of-the-art (SOTA) linear attention architectures. Like AlphaGo's Move 37 that revealed unexpected strategic insights invisible to human players, our AI-discovered architectures demonstrate emergent design principles that systematically surpass human-designed baselines and illuminate previously unknown pathways for architectural innovation. Crucially, we establish the first empirical scaling law for scientific discovery itself--demonstrating that architectural breakthroughs can be scaled computationally, transforming research progress from a human-limited to a computation-scalable process. We provide comprehensive analysis of the emergent design patterns and autonomous research capabilities that enabled these breakthroughs, establishing a blueprint for self-accelerating AI systems.

Due to practical constraints such as partial observability and limited communication, Centralized Training with Decentralized Execution (CTDE) has become the dominant paradigm in cooperative Multi-Agent Reinforcement Learning (MARL). However, existing CTDE methods often underutilize centralized training or lack theoretical guarantees. We propose Multi-Agent Guided Policy Optimization (MAGPO), a novel framework that better leverages centralized training by integrating centralized guidance with decentralized execution. MAGPO uses an auto-regressive joint policy for scalable, coordinated exploration and explicitly aligns it with decentralized policies to ensure deployability under partial observability. We provide theoretical guarantees of monotonic policy improvement and empirically evaluate MAGPO on 43 tasks across 6 diverse environments. Results show that MAGPO consistently outperforms strong CTDE baselines and matches or surpasses fully centralized approaches, offering a principled and practical solution for decentralized multi-agent learning. Our code and experimental data can be found in https://github.com/liyheng/MAGPO.

Designing industrial systems, such as building, improving, and automating distribution centers and manufacturing plants, involves critical decision-making with limited information in the early phases. The lack of information leads to less accurate designs of the systems, which are often difficult to resolve later. It is effective to use simulators to model the designed system and find out the issues early. However, the modeling time required by conventional simulators is too long to allow for rapid model creation to meet decision-making demands. In this paper, we propose a Rapid Modeling Architecture (RMA) for a lightweight industrial simulator that mitigates the modeling burden while maintaining the essential details in order to accelerate and improve decision-making. We have prototyped a simulator based on the RMA and applied it to the actual factory layout design problem. We also compared the modeling time of our simulator to that of an existing simulator, and as a result, our simulator achieved a 78.3% reduction in modeling time compared to conventional simulators.

Building on the conceptual framework presented in our previous work on agentic AI for pharmaceutical research, this paper provides a comprehensive technical analysis of Tippy's multi-agent system implementation for drug discovery laboratory automation. We present a distributed microservices architecture featuring five specialized agents (Supervisor, Molecule, Lab, Analysis, and Report) that coordinate through OpenAI Agents SDK orchestration and access laboratory tools via the Model Context Protocol (MCP). The system architecture encompasses agent-specific tool integration, asynchronous communication patterns, and comprehensive configuration management through Git-based tracking. Our production deployment strategy utilizes Kubernetes container orchestration with Helm charts, Docker containerization, and CI/CD pipelines for automated testing and deployment. The implementation integrates vector databases for RAG functionality and employs an Envoy reverse proxy for secure external access. This work demonstrates how specialized AI agents can effectively coordinate complex laboratory workflows while maintaining security, scalability, reliability, and integration with existing laboratory infrastructure through standardized protocols.

Large Language Model (LLM) agents are increasingly utilized in AI-aided education to support tutoring and learning. Effective communication strategies among LLM agents improve collaborative problem-solving efficiency and facilitate cost-effective adoption in education. However, little research has systematically evaluated the impact of different communication strategies on agents' problem-solving. Our study examines four communication modes, \textit{teacher-student interaction}, \textit{peer-to-peer collaboration}, \textit{reciprocal peer teaching}, and \textit{critical debate}, in a dual-agent, chat-based mathematical problem-solving environment using the OpenAI GPT-4o model. Evaluated on the MATH dataset, our results show that dual-agent setups outperform single agents, with \textit{peer-to-peer collaboration} achieving the highest accuracy. Dialogue acts like statements, acknowledgment, and hints play a key role in collaborative problem-solving. While multi-agent frameworks enhance computational tasks, effective communication strategies are essential for tackling complex problems in AI education.

Recent large-scale events like election fraud and financial scams have shown how harmful coordinated efforts by human groups can be. With the rise of autonomous AI systems, there is growing concern that AI-driven groups could also cause similar harm. While most AI safety research focuses on individual AI systems, the risks posed by multi-agent systems (MAS) in complex real-world situations are still underexplored. In this paper, we introduce a proof-of-concept to simulate the risks of malicious MAS collusion, using a flexible framework that supports both centralized and decentralized coordination structures. We apply this framework to two high-risk fields: misinformation spread and e-commerce fraud. Our findings show that decentralized systems are more effective at carrying out malicious actions than centralized ones. The increased autonomy of decentralized systems allows them to adapt their strategies and cause more damage. Even when traditional interventions, like content flagging, are applied, decentralized groups can adjust their tactics to avoid detection. We present key insights into how these malicious groups operate and the need for better detection systems and countermeasures. Code is available at https://github.com/renqibing/RogueAgent.

The AI alignment problem, which focusses on ensuring that artificial intelligence (AI), including AGI and ASI, systems act according to human values, presents profound challenges. With the progression from narrow AI to Artificial General Intelligence (AGI) and Superintelligence, fears about control and existential risk have escalated. Here, we investigate whether embracing inevitable AI misalignment can be a contingent strategy to foster a dynamic ecosystem of competing agents as a viable path to steer them in more human-aligned trends and mitigate risks. We explore how misalignment may serve and should be promoted as a counterbalancing mechanism to team up with whichever agents are most aligned to human interests, ensuring that no single system dominates destructively. The main premise of our contribution is that misalignment is inevitable because full AI-human alignment is a mathematical impossibility from Turing-complete systems, which we also offer as a proof in this contribution, a feature then inherited to AGI and ASI systems. We introduce a change-of-opinion attack test based on perturbation and intervention analysis to study how humans and agents may change or neutralise friendly and unfriendly AIs through cooperation and competition. We show that open models are more diverse and that most likely guardrails implemented in proprietary models are successful at controlling some of the agents' range of behaviour with positive and negative consequences while closed systems are more steerable and can also be used against proprietary AI systems. We also show that human and AI intervention has different effects hence suggesting multiple strategies.

As large language models (LLMs) are increasingly deployed as autonomous agents, understanding their cooperation and social mechanisms is becoming increasingly important. In particular, how LLMs balance self-interest and collective well-being is a critical challenge for ensuring alignment, robustness, and safe deployment. In this paper, we examine the challenge of costly sanctioning in multi-agent LLM systems, where an agent must decide whether to invest its own resources to incentivize cooperation or penalize defection. To study this, we adapt a public goods game with institutional choice from behavioral economics, allowing us to observe how different LLMs navigate social dilemmas over repeated interactions. Our analysis reveals four distinct behavioral patterns among models: some consistently establish and sustain high levels of cooperation, others fluctuate between engagement and disengagement, some gradually decline in cooperative behavior over time, and others rigidly follow fixed strategies regardless of outcomes. Surprisingly, we find that reasoning LLMs, such as the o1 series, struggle significantly with cooperation, whereas some traditional LLMs consistently achieve high levels of cooperation. These findings suggest that the current approach to improving LLMs, which focuses on enhancing their reasoning capabilities, does not necessarily lead to cooperation, providing valuable insights for deploying LLM agents in environments that require sustained collaboration. Our code is available at https://github.com/davidguzmanp/SanctSim

Estimating high-dimensional covariance matrices is a key task across many fields. This paper explores the theoretical limits of distributed covariance estimation in a feature-split setting, where communication between agents is constrained. Specifically, we study a scenario in which multiple agents each observe different components of i.i.d. samples drawn from a sub-Gaussian random vector. A central server seeks to estimate the complete covariance matrix using a limited number of bits communicated by each agent. We obtain a nearly tight minimax lower bound for covariance matrix estimation under operator norm and Frobenius norm. Our main technical tool is a novel generalization of the strong data processing inequality (SDPI), termed the Conditional Strong Data Processing Inequality (C-SDPI) coefficient, introduced in this work. The C-SDPI coefficient shares key properties such as tensorization with the conventional SDPI. Crucially, it quantifies the average contraction in a state-dependent channel and can be significantly lower than the worst-case SDPI coefficient over the state input. Utilizing the doubling trick of Geng-Nair and an operator Jensen inequality, we compute this coefficient for Gaussian mixture channels. We then employ it to establish minimax lower bounds on estimation error, capturing the trade-offs among sample size, communication cost, and data dimensionality. Building on this, we present a nearly optimal estimation protocol whose sample and communication requirements match the lower bounds up to logarithmic factors. Unlike much of the existing literature, our framework does not assume infinite samples or Gaussian distributions, making it broadly applicable. Finally, we extend our analysis to interactive protocols, showing interaction can significantly reduce communication requirements compared to non-interactive schemes.

Central to agentic capability and trustworthiness of language model agents (LMAs) is the extent they maintain stable, reliable, identity over time. However, LMAs inherit pathologies from large language models (LLMs) (statelessness, stochasticity, sensitivity to prompts and linguistically-intermediation) which can undermine their identifiability, continuity, persistence and consistency. This attrition of identity can erode their reliability, trustworthiness and utility by interfering with their agentic capabilities such as reasoning, planning and action. To address these challenges, we introduce \textit{agent identity evals} (AIE), a rigorous, statistically-driven, empirical framework for measuring the degree to which an LMA system exhibit and maintain their agentic identity over time, including their capabilities, properties and ability to recover from state perturbations. AIE comprises a set of novel metrics which can integrate with other measures of performance, capability and agentic robustness to assist in the design of optimal LMA infrastructure and scaffolding such as memory and tools. We set out formal definitions and methods that can be applied at each stage of the LMA life-cycle, and worked examples of how to apply them.