Recent advances in large language models (LLMs) have opened new opportunities for academic literature retrieval. However, existing systems often rely on rigid pipelines and exhibit limited reasoning capabilities. We introduce SPAR, a multi-agent framework that incorporates RefChain-based query decomposition and query evolution to enable more flexible and effective search. To facilitate systematic evaluation, we also construct SPARBench, a challenging benchmark with expert-annotated relevance labels. Experimental results demonstrate that SPAR substantially outperforms strong baselines, achieving up to +56% F1 on AutoScholar and +23% F1 on SPARBench over the best-performing baseline. Together, SPAR and SPARBench provide a scalable, interpretable, and high-performing foundation for advancing research in scholarly retrieval. Code and data will be available at: https://github.com/xiaofengShi/SPAR

Traffic Signal Control (TSC) is essential for managing urban traffic flow and reducing congestion. Reinforcement Learning (RL) offers an adaptive method for TSC by responding to dynamic traffic patterns, with multi-agent RL (MARL) gaining traction as intersections naturally function as coordinated agents. However, due to shifts in environmental dynamics, implementing MARL-based TSC policies in the real world often leads to a significant performance drop, known as the sim-to-real gap. Grounded Action Transformation (GAT) has successfully mitigated this gap in single-agent RL for TSC, but real-world traffic networks, which involve numerous interacting intersections, are better suited to a MARL framework. In this work, we introduce JL-GAT, an application of GAT to MARL-based TSC that balances scalability with enhanced grounding capability by incorporating information from neighboring agents. JL-GAT adopts a decentralized approach to GAT, allowing for the scalability often required in real-world traffic networks while still capturing key interactions between agents. Comprehensive experiments on various road networks under simulated adverse weather conditions, along with ablation studies, demonstrate the effectiveness of JL-GAT. The code is publicly available at https://github.com/DaRL-LibSignal/JL-GAT/.

Multi-agent systems (MASs) consisting of a number of autonomous agents that communicate, coordinate, and jointly sense the environment to achieve complex missions can be found in a variety of applications such as robotics, smart cities, and internet-of-things applications. Modeling and monitoring MAS requirements to guarantee overall mission objectives, safety, and reliability is an important problem. Such requirements implicitly require reasoning about diverse sensing and communication modalities between agents, analysis of the dependencies between agent tasks, and the spatial or virtual distance between agents. To capture such rich MAS requirements, we model agent interactions via multiple directed graphs, and introduce a new logic -- Spatio-Temporal Logic with Graph Operators (STL-GO). The key innovation in STL-GO are graph operators that enable us to reason about the number of agents along either the incoming or outgoing edges of the underlying interaction graph that satisfy a given property of interest; for example, the requirement that an agent should sense at least two neighboring agents whose task graphs indicate the ability to collaborate. We then propose novel distributed monitoring conditions for individual agents that use only local information to determine whether or not an STL-GO specification is satisfied. We compare the expressivity of STL-GO against existing spatio-temporal logic formalisms, and demonstrate the utility of STL-GO and our distributed monitors in a bike-sharing and a multi-drone case study.

While human infants robustly discover their own causal efficacy, standard reinforcement learning agents remain brittle, as their reliance on correlation-based rewards fails in noisy, ecologically valid scenarios. To address this, we introduce the Causal Action Influence Score (CAIS), a novel intrinsic reward rooted in causal inference. CAIS quantifies an action's influence by measuring the 1-Wasserstein distance between the learned distribution of sensory outcomes conditional on that action, $p(h|a)$, and the baseline outcome distribution, $p(h)$. This divergence provides a robust reward that isolates the agent's causal impact from confounding environmental noise. We test our approach in a simulated infant-mobile environment where correlation-based perceptual rewards fail completely when the mobile is subjected to external forces. In stark contrast, CAIS enables the agent to filter this noise, identify its influence, and learn the correct policy. Furthermore, the high-quality predictive model learned for CAIS allows our agent, when augmented with a surprise signal, to successfully reproduce the "extinction burst" phenomenon. We conclude that explicitly inferring causality is a crucial mechanism for developing a robust sense of agency, offering a psychologically plausible framework for more adaptive autonomous systems.

The future of software engineering--SE 3.0--is unfolding with the rise of AI teammates: autonomous, goal-driven systems collaborating with human developers. Among these, autonomous coding agents are especially transformative, now actively initiating, reviewing, and evolving code at scale. This paper introduces AIDev, the first large-scale dataset capturing how such agents operate in the wild. Spanning over 456,000 pull requests by five leading agents--OpenAI Codex, Devin, GitHub Copilot, Cursor, and Claude Code--across 61,000 repositories and 47,000 developers, AIDev provides an unprecedented empirical foundation for studying autonomous teammates in software development. Unlike prior work that has largely theorized the rise of AI-native software engineering, AIDev offers structured, open data to support research in benchmarking, agent readiness, optimization, collaboration modeling, and AI governance. The dataset includes rich metadata on PRs, authorship, review timelines, code changes, and integration outcomes--enabling exploration beyond synthetic benchmarks like SWE-bench. For instance, although agents often outperform humans in speed, their PRs are accepted less frequently, revealing a trust and utility gap. Furthermore, while agents accelerate code submission--one developer submitted as many PRs in three days as they had in three years--these are structurally simpler (via code complexity metrics). We envision AIDev as a living resource: extensible, analyzable, and ready for the SE and AI communities. Grounding SE 3.0 in real-world evidence, AIDev enables a new generation of research into AI-native workflows and supports building the next wave of symbiotic human-AI collaboration. The dataset is publicly available at https://github.com/SAILResearch/AI_Teammates_in_SE3. > AI Agent, Agentic AI, Coding Agent, Agentic Coding, Software Engineering Agent

The detection of outliers within cryptocurrency limit order books (LOBs) is of paramount importance for comprehending market dynamics, particularly in highly volatile and nascent regulatory environments. This study conducts a comprehensive comparative analysis of robust statistical methods and advanced machine learning techniques for real-time anomaly identification in cryptocurrency LOBs. Within a unified testing environment, named AITA Order Book Signal (AITA-OBS), we evaluate the efficacy of thirteen diverse models to identify which approaches are most suitable for detecting potentially manipulative trading behaviours. An empirical evaluation, conducted via backtesting on a dataset of 26,204 records from a major exchange, demonstrates that the top-performing model, Empirical Covariance (EC), achieves a 6.70% gain, significantly outperforming a standard Buy-and-Hold benchmark. These findings underscore the effectiveness of outlier-driven strategies and provide insights into the trade-offs between model complexity, trade frequency, and performance. This study contributes to the growing corpus of research on cryptocurrency market microstructure by furnishing a rigorous benchmark of anomaly detection models and highlighting their potential for augmenting algorithmic trading and risk management.

Collaboration among multiple large language model (LLM) agents is a promising approach to overcome inherent limitations of single-agent systems, such as hallucinations and single points of failure. As LLM agents are increasingly deployed on open blockchain platforms, multi-agent systems capable of tolerating malicious (Byzantine) agents have become essential. Recent Byzantine-robust multi-agent systems typically rely on leader-driven coordination, which suffers from two major drawbacks. First, they are inherently vulnerable to targeted attacks against the leader. If consecutive leaders behave maliciously, the system repeatedly fails to achieve consensus, forcing new consensus rounds, which is particularly costly given the high latency of LLM invocations. Second, an underperforming proposal from the leader can be accepted as the final answer even when higher-quality alternatives are available, as existing methods finalize the leader's proposal once it receives a quorum of votes. To address these issues, we propose DecentLLMs, a novel decentralized consensus approach for multi-agent LLM systems, where worker agents generate answers concurrently and evaluator agents independently score and rank these answers to select the best available one. This decentralized architecture enables faster consensus despite the presence of Byzantine agents and consistently selects higher-quality answers through Byzantine-robust aggregation techniques. Experimental results demonstrate that DecentLLMs effectively tolerates Byzantine agents and significantly improves the quality of selected answers.

Large-language-model (LLM) agents exhibit complex, context-sensitive behaviour that quickly renders static benchmarks and ad-hoc manual testing obsolete. We present Neo, a configurable, multi-agent framework that automates realistic, multi-turn evaluation of LLM-based systems. Neo couples a Question Generation Agent and an Evaluation Agent through a shared context-hub, allowing domain prompts, scenario controls and dynamic feedback to be composed modularly. Test inputs are sampled from a probabilistic state model spanning dialogue flow, user intent and emotional tone, enabling diverse, human-like conversations that adapt after every turn. Applied to a production-grade Seller Financial Assistant chatbot, Neo (i) uncovered edge-case failures across five attack categories with a 3.3% break rate close to the 5.8% achieved by expert human red-teamers, and (ii) delivered 10-12X higher throughput, generating 180 coherent test questions in around 45 mins versus 16h of human effort. Beyond security probing, Neo's stochastic policies balanced topic coverage and conversational depth, yielding broader behavioural exploration than manually crafted scripts. Neo therefore lays a foundation for scalable, self-evolving LLM QA: its agent interfaces, state controller and feedback loops are model-agnostic and extensible to richer factual-grounding and policy-compliance checks. We release the framework to facilitate reproducible, high-fidelity testing of emerging agentic systems.

Whole slide images (WSIs) are vital in digital pathology, enabling gigapixel tissue analysis across various pathological tasks. While recent advancements in multi-modal large language models (MLLMs) allow multi-task WSI analysis through natural language, they often underperform compared to task-specific models. Collaborative multi-agent systems have emerged as a promising solution to balance versatility and accuracy in healthcare, yet their potential remains underexplored in pathology-specific domains. To address these issues, we propose WSI-Agents, a novel collaborative multi-agent system for multi-modal WSI analysis. WSI-Agents integrates specialized functional agents with robust task allocation and verification mechanisms to enhance both task-specific accuracy and multi-task versatility through three components: (1) a task allocation module assigning tasks to expert agents using a model zoo of patch and WSI level MLLMs, (2) a verification mechanism ensuring accuracy through internal consistency checks and external validation using pathology knowledge bases and domain-specific models, and (3) a summary module synthesizing the final summary with visual interpretation maps. Extensive experiments on multi-modal WSI benchmarks show WSI-Agents's superiority to current WSI MLLMs and medical agent frameworks across diverse tasks.

Popular methods in cooperative Multi-Agent Reinforcement Learning with partially observable environments typically allow agents to act independently during execution, which may limit the coordinated effect of the trained policies. However, by sharing information such as known or suspected ongoing threats, effective communication can lead to improved decision-making in the cyber battle space. We propose a game design where defender agents learn to communicate and defend against imminent cyber threats by playing training games in the Cyber Operations Research Gym, using the Differentiable Inter Agent Learning algorithm adapted to the cyber operational environment. The tactical policies learned by these autonomous agents are akin to those of human experts during incident responses to avert cyber threats. In addition, the agents simultaneously learn minimal cost communication messages while learning their defence tactical policies.