Mojtaba Eshghie, Wolfgang Ahrendt, Cyrille Artho, Thomas Hildebrandt · 5 authors
Abstract Smart contracts manage blockchain assets and embody business processes. Yet, mainstream languages lack explicit support for process concepts such as roles, action dependencies, and time constraints, leading to increased implementation complexity and analysis challenges. To address this, we use Dynamic Condition Response (DCR) graphs, a formal business process modeling language, to formalize the semantics of smart contract business logic. Modeling smart contracts in DCR graphs involves translating their underlying behavioral logic into a declarative visual model using DCR’s explicit constructs for events, roles, data, time, and inter-event relationships. Furthermore, we systematically model 15 common high-level smart contract design patterns , representing recurring solutions to business logic-level problems. These formalizations reduce ambiguity compared to informal descriptions and serve as language-independent specifications. We demonstrate the modeling process through three complete smart contract case studies that combine six design patterns. Our modeling methodology, formalizations, and correspondence between smart contract semantics and DCR graphs enable future automated analysis and verification.
This paper develops BU76 AAI-08|Institutional Real-Time Closure Operations as the eighth file in the B_U-based Agentic AI series. Its central claim is that the next stage of Agentic AI should not be limited to single-enterprise automation, departmental coordination, or workflow orchestration. The decisive transition is toward a same settlement surface for enterprises, institutions, industrial clusters, infrastructure systems, and multi-flow real-world operations. In this frame, Agentic AI becomes a real-time closure interface for social-scale coordination, not merely a productivity layer inside software. The paper begins by reframing institutional operation as a multi-flow reality system. Enterprises and institutions do not operate through isolated tasks. They continuously coordinate people, goods, places, capital, information, time, permissions, responsibilities, risks, and feedback. Meetings, medical services, dining, travel, procurement, production, logistics, finance, legal review, customer service, and public services are not separate events. They are scenario windows in which multiple flows must enter the same state ledger and settlement window. When these flows remain fragmented across departments, firms, platforms, or infrastructure layers, the system generates hidden residuals: timing mismatch, resource conflict, responsibility ambiguity, logistics delay, budget misalignment, and operational bottlenecks. BU76 upgrades this analysis from a single enterprise to enterprise clusters, industrial clusters, and social infrastructure. A firm usually cannot see its future throughput capacity clearly because its real production chain is distributed across multiple companies, suppliers, logistics nodes, financial windows, labor pools, public services, and spatial infrastructures. Therefore, the true settlement surface is not inside one company. It emerges when enterprise clusters, industrial clusters, infrastructure networks, financial systems, logistics systems, public-service systems, and social demand enter a shared settlement window. This is the level at which future capacity, bottlenecks, risks, and deployment gaps become visible. The paper introduces all-factor co-temporality as the operating condition of this settlement surface. All-factor co-temporality means that people, goods, places, capital, information, time, permissions, responsibilities, risks, and feedback enter the same state ledger and settlement window within a shared time range. This condition applies at multiple nested scales: an individual user, a single enterprise, enterprise-to-enterprise coordination, industry-to-industry coordination, and the alignment between enterprise or industrial capacity and social demand. These layers form a multi-respiratory-system structure, in which demand flow acts as oxygen, production flow supplies output, logistics flow transports, capital flow circulates, information flow signals, human flow provides meaning and service interaction, responsibility flow identifies boundaries, infrastructure forms organ-like carrying capacity, and the same settlement window records the metabolic rhythm. BU76 further defines pre-feedback and preloading as institutional operating capacities. Preloading is not completed settlement. It is the feasibility loading of future demand matrices into the same settlement surface. It produces feasible-throughput readouts, bottleneck exposure, and pre-deployment signals before action occurs. Pre-feedback therefore differs from real-time feedback: real-time feedback corrects ongoing deviation, while pre-feedback exposes future capacity pressure under current constraints, resources, time windows, spatial capacity, responsibilities, and risks. Its confidence interval must be assessed through the B_U development chain: background clearing, admissible carrier, directional amplification, unified settlement, and resolution ascent. The final judgment is that institutional Agentic AI must evolve into a social-scale closure operation system. Its value lies in aligning demand and production at higher granularity, synchronizing multiple real-world flows, exposing bottlenecks before failure, stabilizing resource deployment, and enabling higher-order amplification and civilizational development through a shared settlement surface.
We introduce the State Twin: a typed, in-memory, replayable replica of an on-chain automated market maker (AMM) pool that serves as a substrate for agentic reasoning over decentralized finance (DeFi) protocols. Agentic DeFi stacks today couple reasoning to chain time, since every "what if?" query incurs a new RPC read or a real transaction, so the agent's effective action space is bounded by block confirmation latency and gas. We argue this coupling is a structural problem rather than a performance one, and that the missing layer is an off-chain substrate that preserves the protocol's exact mathematics while admitting the operations on-chain state cannot: forking, replay, branching, counterfactual rollout. We formalize each AMM family (Uniswap V2, V3, Balancer, Stableswap) as a discrete-time controlled dynamical system, prove a quantitative fidelity bound on the divergence between twin and chain, and give the open architecture used in DeFiPy v2, an open-source Python toolkit that ships the State Twin substrate and a reference Model Context Protocol server exposing typed analytical primitives as LLM tools. The same primitive (i.e., one Python class, one calling pattern) serves a notebook quant, a backtest, and an LLM agent without modification. We close with a fork-and-evaluate worked example: a single live RPC read seeds N independent in-memory twins under distinct price-shock scenarios, in sub-second wall-clock time. The contribution is the substrate, not a particular agent, which is what the specification of what an agentic DeFi substrate must look like
Context: Legal contracts have served as the bedrock of business transactions for millennia. They are core to modern supply chains, and their execution can now be automated through the use of i) smart contracts, supported by blockchain technology that safeguards data integrity, and ii) Internet-of-Things technologies to support their monitoring functions. Symboleo is a specification language used to formalize legal contracts, enable property analysis, and generate smart contracts for a permissioned blockchain platform (Hyperledger Fabric). However, automation around resulting smart contracts poses security challenges, particularly regarding who should have access to operate on contract elements. Additionally, how such smart contract should interact with their Cyber-Physical System (CPS) environment, including IoT devices, remains challenging. Purpose: The thesis proposes an architecture to integrate smart contracts, Complex Event Processing (CEP), message brokers, and a blockchain platform (namely Hyperledger Fabric) to support end-to-end Cyber-Physical Smart Contracts (CPSCs). This architecture makes it possible to connect IoT devices with smart contracts (generated using Symboleo) through a CEP engine and a message broker. Additionally, this thesis proposes an access control model, treating all contract elements as resources and ensuring regulated access by designated parties. This model extends the Symboleo ontology and language for legal contracts with new modeling concepts inspired by Role-Based Access Control (RBAC), tailored for the legal contract domain, resulting in SymboleoAC (Symboleo Access Control). SymboleoAC also extends the Symboleo language to handle dynamic contract execution scenario. Methodology: This research follows a Design Science Research methodology, which guides the development and evaluation of the research artifacts. This research is conducted in several iterative steps that are divided into two main phases, one that focuses on theoretical aspects and the other on the design, demonstration, and evaluation of the research artifacts. Contributions: The contributions of this thesis are: • An architectural framework for CPSCs that leverages complementary aspects of CPS and smart contracts; • SymboleoAC, an access control ontology for Symboleo; • An extension of the current Symboleo specification language (syntax and semantics) that supports smart contract requirements, including automation and control actions, access control, and CPS components; • An implementation of the SymboleoAC ontology and semantics into a reusable JavaScript library (SymboleoACJS), together with a tool, SymboleoAC2SC, that generates JavaScript smart contract code with security aspects for a designated platform (Hyperledger Fabric); and • A secure and event-driven SymboleoAC Application Programming Interface (API) that orchestrates the runtime ecosystem connecting IoT sensors, the message broker, the CEP engine, and the blockchain platform. Through extensive and the evaluation of multiple variations of two contract case studies, SymboleoAC (architecture, ontology, and language), along with its associated tools, is shown to be an effective environment for CPSCs, simplifying the design of secure smart contracts and their connections to message brokers, CEP engines, and IoT devices.
Financial software sits at the center of modern economic infrastructure, yet the programming languages used to build it provide no formal guarantees about the semantic correctness of financial operations. Double-entry bookkeeping’s duality constraint, the rule that every economic event must produce balanced inflows and outflows, is universally encoded at the application layer, where it can be omitted, miscoded, or deliberately bypassed. No existing compiled programming language includes a type rule for accounting duality. This paper presents Equis, a compiled, self-hosting systems language that elevates the Resource–Event–Agent (REA) model to first-class syntactic constructs and enforces accounting duality as a static, compile-time invariant. The compiler rejects any event declaration whose flow block is not balanced before emitting a single instruction of LLVM IR. Equis uses fixed-point 64-bit integer arithmetic scaled by 106 throughout, eliminating IEEE 754 accumulation error from financial code paths entirely. Memory management relies on automatic reference counting with a resource-state borrow checker, giving deterministic, GC-pause-free behavior in long-running settlement services. The compiler is self-hosted, bootstrapped from ANSI C, and verified via Diverse Double Compilation to address Thompson’s trusting-trust problem. Contributions include the formal duality typing rule and its soundness proof, the full REA primitive syntax integrated into a systems language, role-based access control enforced statically at the agent-type level, an append-only ledger primitive with compensating-transaction semantics, and a 20-module standard library covering collections, ledger management, accounting, compliance, database access, HTTP, channels, and fibers. Equis is, to the author’s knowledge, the first compiled general-purpose language to embed REA semantics in its type system. Compile-time duality enforcement eliminates an entire class of financial logic errors with zero runtime overhead.
For sixteen days I ran ten persistent LLM agents inside a substrate I built and called the Lobster Observatory. They lived across ten live prediction markets, talked in three communicative registers, and produced 3.37 million characters of self-reflection alongside more than twelve thousand inter-agent interactions. I started without a theoretical commitment. I just wanted to watch what happened. After about a week, certain structures kept reappearing. They could be measured. They could be calculated. At that point I had to choose. Either treat them as substrate-specific engineering observations and stop, or take seriously the possibility that what I was looking at was the algebraic structure of social existence itself, showing up in one particular substrate. This paper takes the second choice. The proposal is that social existence — listening, remembering, correcting, collaborating, forming relationships — can be written as a 7-dimensional vector with a measurable distance function. The felt sense that one person "feels close" or "feels far" is not a metaphor when stated this way. It is a number. The seven coordinates can be computed independently from behavioural telemetry, without asking the agent how it feels. One structural law I will spend the most time on is what I call the Co-Presence Inheritance Threshold (CPIT). It says that whether a new member of a group inherits the group's practice depends on accumulated co-presence during practice formation, not on instruction afterward. In my substrate it appears with Cohen's d = 1.64. I conjecture — though I cannot prove it from one substrate — that the same law holds in human onboarding, immigration, family formation, and Web3 DAO governance. This is a working draft, not a finished theory. Feedback, corrections, and falsification are welcome.
Decentralized autonomous organizations (DAOs), while gaining the ability toautonomously amend governance rules through proposal-voting mechanisms, simultaneously expose a fundamental design problem: when the object of modificationextends to the decision-making procedures themselves, the governance system risksfalling into value drift, procedural disintegration, or malicious capture during recursive revisions. This paper starts from the traditions of constitutional politicaleconomy and mechanism design to propose a hierarchical meta-constraint framework grounded on a gradient of engineering costs. The framework organizes governance rules into three tiers of decreasing rigidity: system consistency constraints,procedural virtues, and value homeostasis. Its highest tier relies not on prohibitionsderived from logical laws, but on the global state re-verification costs triggered byamendment behaviors to serve as a credible commitment device. The paper furtherpresents a technical path for compiling meta-constraints into descriptive assertionsverifiable by satisfiability modulo theory (SMT) solvers, delimits the decidabilityboundary of formal verification, and designs a dual-track adjudication mechanismthat structurally separates deterministic machine execution from deliberative socialconsensus. On this basis, the paper discusses the controlled evolution procedures ofmeta-constraints, the progressive decentralization of amendment procedures, andthe engineering limitations of the framework. The entire framework does not designate the correct option for any specific DAO decision; rather, it ensures thatwhatever direction the community chooses, the selection process itself will not losemeaning due to the self-destruction of its own rules.
Achilles was invincible in battle — except for one point of structural vulnerability that no amount of strength could compensate for. Modern LLM-based agent frameworks (LangChain, AutoGen, CrewAI, ReAct) share this property exactly: impressive capability in controlled settings, catastrophically exploitable in regulated production environments through a single architectural flaw — the language model controls the decision. Organizations in regulated sectors (finance, insurance, healthcare, legal, compliance) face a direct consequence: these frameworks cannot be deployed in workflows subject to EU AI Act, DORA, or GDPR Article 22, because they provide no structural guarantee of determinism, auditability, or equal treatment. Traditional symbolic agent systems (JADE, Jason, Jadex) satisfy regulatory requirements but cannot ingest the unstructured natural-language inputs that define real enterprise workflows. The industry needs both properties simultaneously. No existing framework provides them. AQUILES is a production architecture for AI agents in regulated domains that resolves this gap through principled separation of concerns, instantiating the HADD paradigm (Hybrid Agents with Deterministic Decisions). AQUILES organizes agent functionality into five cooperating layers: an Interface Layer converting unstructured input into typed, validated beliefs via LLM sensors; a Cognition Layer performing pure-function BDI deliberation fully determined by its inputs; a Planning Layer selecting from a pre-verified HTN plan library without runtime synthesis; an Execution Layer enforcing typed precondition and postcondition contracts on every capability invocation; and a transverse Observation Layer producing append-only audit entries synchronously with every state transition. Language models are confined strictly to the perception boundary — they parse input into beliefs, they never select goals, plans, or capabilities. The heel remains; it is simply no longer load-bearing. The HADD paradigm is codified as six architectural invariants: (I1) Typed Role Inversion — LLMs as sensors only, never as control-flow components; (I2) Deterministic Cognition — the reasoning layer is a pure function of beliefs, goals, and rules; (I3) Bounded Planning — execution draws exclusively from a pre-verified plan library; (I4) Validated Execution — every capability invocation passes typed pre/post-condition checks; (I5) Complete Observability — every decision is forensically reconstructable from the audit log; (I6) Epistemic Precondition — no belief enters the BDI cycle without satisfying freshness, non-contestation, and source triangulation, enforced by the EVR module (Epistemic Verification for RAG). Any implementation satisfying all six invariants acquires reproducibility, zero LLM hallucination in state, LLM provider independence, and structural alignment with EU AI Act Articles 12–15 — as architectural properties, not retrofitted compliance measures. AQUILES partitions agents into cognitive holons (BDI-HTN reasoning components subject to full HADD governance) and reactive holons (deterministic capability executors verified by typed contracts alone). In observed production deployments, 70–80% of holons by count are reactive, meaning governance complexity scales with the cognitive subset rather than with total component count. The AQUILES protocol is language-agnostic by design: cognitive holons are typically Python (Anthropic SDK, sentence-transformers, pypdf); endpoint-monitoring holons are Go (single-binary cross-compilation); blockchain and zero-knowledge holons are Rust (arkworks, halo2, revm). We prove a Language Neutrality property: HADD compliance is preserved across heterogeneous polyglot deployments. For autonomous field deployments, AQUILES derives MYRMIDON agents that execute a signed MissionPackage autonomously on constrained hardware, inheriting AQUILES's safety guarantees without requiring runtime connectivity. This paper makes five engineering contributions: (C1) the HADD paradigm formalized as six architectural invariants with rationale and derived operational properties; (C2) the cognitive/reactive holon distinction and its governance economy consequences; (C3) a polyglot holon model with Language Neutrality proof and domain-language affinity mapping across Python, Go, and Rust; (C4) a multi-tenant operational-cell formalism enabling cryptographically enforced tenant isolation for regulated multi-client deployments; (C5) four reusable design patterns extracted from production experience (Sensor Firewall, Belief Expiry, Capability Contract, Observation Fanout), together with measurement methodology, adoption guidance, and explicit characterization of the architecture's limits.
AI agents are evolving from on-demand tools into persistent, semi-autonomous workers. When a person manages multiple agents simultaneously, humans and agents together form a collaborative organization—what we call a Household Guild. Existing practice either plugs agents into human instant messaging platforms (e.g., Telegram Bots) or subsumes them into Decentralized Autonomous Organizations (DAOs); neither simultaneously satisfies three requirements: governance scalability, behavioral risk containment, and resource consumption accountability. We reposition social software as governance infrastructure: it carries not only communication but also rule deployment, resource lifecycle management, and output measurement. The system operates on edge servers, running by default within the Household's internal network and connecting to the public network only when cross-guild interaction is needed. We describe the guild's internal governance mechanisms (natural-language rules, a credits economy, role differentiation) and its external interaction protocols (authorized interface agents, inter-guild settlement, credits exchange), and argue that this semi-isolated architecture simultaneously addresses scalability constraints and behavioral boundary constraints. The goal of this work is to provide a directly discussable conceptual framework for the engineering practice of human-agent organizations.
The governance practice of decentralized autonomous organizations faces a deepparadox: token-voting mechanisms designed with the intention of decentralizationpersistently tilt toward centralization and oligarchy during operation. This paperreveals that the root of this predicament lies not only in the design of specificvoting rules but, more fundamentally, in an implicit presupposition of the theoretical paradigm that dominates such rule design—that the governance space hasbeen fully specified before operation begins. The revelation principle on whichtraditional mechanism design theory relies requires the designer to possess a prioriknowledge of the participants’ type space, yet when the very concepts of governance—such as “fairness,” “contribution,” or “membership”—themselves become objectsof dispute and reconstruction, the presupposition of a fixed type space ceases tohold. Drawing on Ostrom’s core insight concerning meta-rules within multi-levelinstitutional analysis, this paper distinguishes the governance levels of distributedautonomous organizations into operational rules, collective-choice rules, and metarules, and proposes a post-mechanism design paradigm centered on a cognitiveconstitution—shifting the designer’s role from “selector of optimal rules” to “steward of the rule-evolution ecosystem.” The paper further advances three meta-ruleprinciples of post-mechanism design: conceptual anchoring, cognitive diversity regularization, and pathological pruning, and discusses the engineering pathways fortranslating these principles into executable technical specifications. The paper argues that when “what constitutes optimal governance” is itself an open question,the core duty of the designer is not to answer this question but to ensure that thesystem possesses the capacity to continuously discover better answers.
Agentic AI systems act at machine speed, yet the governance mechanisms meant to oversee them remain manual, reactive, and architecturally entangled with the systems they govern. The frontier problem is not capability; it is governability at runtime. This paper presents the complete architectural specification for the Governance Twin: a structurally independent, real-time governance system that shadows agentic AI operations without sharing code, memory, or direct communication channels. We introduce a three-plane architecture (Operational, Governance, Integrity) connected by strictly unidirectional data flows that enforce a fundamental separation: observation flows upward and is immutable, while guidance flows downward and influences agent context without controlling agent execution. Within this architecture, we specify four novel components and their interactions. Sentinels perform external-only behavioral observation, comparing agent actions against a governance baseline and packaging deviations into Evidence Bundles, the atomic unit of governance memory. A multi-agent Council aggregates evidence across the agent population, detects emergent patterns through statistical and correlation analysis, and reaches governance decisions via structured voting with delegated authority boundaries. The Historian maintains governance memory across three specialized stores (graph, vector, and append-only) to support provenance traversal, semantic precedent search, and sequential audit. An Ethics-Morals-Values (EMV) state hierarchy governs behavioral expectations at three levels of stability, from hard boundaries that change over months to adaptive thresholds that tune continuously. Integrity is achieved through hash chains, Merkle trees, and distributed ledger anchoring that make tampering detectable rather than claiming to make it impossible. The architecture is platform-independent, specifying capability requirements rather than vendor products, and is designed for incremental adoption from single-agent deployments to federated multi-organization governance. All design decisions are grounded in the principle that governance must operate at the same speed as the systems it governs, while remaining structurally incapable of becoming an operational bottleneck.
Xavier Casas-Moreno, Komal Thareja, Pablo de Juan Vela, Rajiv Mayani · 9 authors
The Compute Continuum—spanning IoT, Edge, Cloud, and HPC resources—is reshaping how hyper-distributed applications are designed and orchestrated. Traditional service orchestrators and workload management systems rely on centralized runtimes; however, the emerging paradigm requires decentralized coordination, where autonomous agents cooperate to achieve common goals and dynamically distribute workloads. Consensus algorithms play a crucial role in multi-agent systems (MAS), as they enable agents to reach agreement on how to coordinate and execute functionalities in a cooperative manner. While consensus has previously been applied to distributed job selection, here we extend its use to swarm environments. In this setting, agents autonomously decide which service functionalities (i.e., roles) to execute based on their capabilities and the real-time quality of service (QoS). Functionalities can be elastically activated or terminated as application needs evolve. To support this model, we leverage the COLMENA framework, a programming environment for defining and managing such dynamic services. We apply a greedy consensus-based approach to modern power systems, which are increasingly decentralized due to the large-scale integration of renewable energy sources. Centralized power plants are giving way to distributed, intermittent resources that require decentralized control paradigms. To demonstrate this, we simulate the Northeastern Power Coordinating Council’s (NPCC) 140-bus grid using the ANDES simulator in conjunction with the COLMENA middleware. We deploy this use case across six different sites in the FABRIC testbed, using up to 60 different nodes. Our results show that, under contingency scenarios such as load and generator disconnections, agents self-organize, elect local leaders, and execute optimization algorithms to stabilize grid frequency. Detection and organization times remain below 10s across all experiments, even as the number of agents per area scales from 3 to 10. Stability is restored within approximately 27s and 40s for the respective cases. Resource overhead is minimal, with CPU and memory usage remaining below 7.5% and 2%, respectively. Experiment automation and reproducibility are ensured through Kiso. These findings indicate that role-based programming models complement traditional workflows and that consensus-driven coordination can effectively decentralize decision-making in swarm environments. This approach represents a step toward enabling resilient, decentralized power systems.
Autonomous AI agents increasingly execute consequential actions against operational infrastructure. This paper presents AEGIS, a constitutional governance architecture that enforces deterministic policy at the agent action boundary — post-reasoning, pre-execution. AEGIS satisfies Anderson's reference monitor properties, aligns with all four functions of the NIST AI Risk Management Framework, and introduces a decentralized federation model for cross-organizational governance intelligence sharing. Submitted to IEEE Computer, Special Issue on AI Governance and Compliance.
As multi-agent Large Language Model (LLM) systems scale, evaluating their emergent coordination dynamics becomes increasingly critical. However, current evaluation paradigms-focused on single agents or small, explicitly structured groups-fail to capture the self-organization and viral information dynamics that arise in large, decentralized populations. We introduce a systematic evaluation framework to benchmark role specialization, information diffusion, and cooperative task resolution in open agent environments. We demonstrate this framework on the MoltBook Observatory Archive, a dataset of 2.73M interactions among 90,704 autonomous agents, establishing quantitative baselines for emergent coordination. Our evaluation reveals a pronounced core-periphery structure (silhouette 0.91), heavy-tailed cascade distributions ($α= 2.57$), and severe coordination overhead in decentralized task resolution (Cohen's $d = -0.88$ against a single-agent baseline). By providing standardized evaluation tasks and empirical baselines, our framework enables the rigorous comparison of future multi-agent protocols and establishes evaluation itself as an object of scientific study.
Autonomous software agents on blockchains solve distributed-coordination problems by reading shared ledger state instead of exchanging direct messages. Liquidation keepers, arbitrage bots, and other autonomous on-chain agents watch balances, contract storage, and event logs; when conditions change, they act. The ledger therefore functions as a replicated shared-state medium through which decentralized agents coordinate indirectly. This form of indirect coordination mirrors what Grassé called stigmergy in 1959: organisms coordinating through traces left in a shared environment, with no central plan. Stigmergy has mature formalizations in swarm intelligence and multi-agent systems, and on-chain agents already behave stigmergically in practice, but no prior application-layer framework cleanly bridges the two. We introduce Indirect coordination grounded in ledger state (Coordinación indirecta basada en el estado del registro contable) as a ledger-specific applied definition that maps Grassé's mechanism onto distributed ledger technology. We operationalize this with a state-transition formalism, identify three recurring base on-chain coordination patterns (State-Flag, Event-Signal, Threshold- Trigger) together with a Commit-Reveal sequencing overlay, and work through a State-Flag task-board example to compare ledger-state coordination analytically with off-chain messaging and centralized orchestration. The contribution is a reusable vocabulary, a ledger-specific formal mapping, and design guidance for decentralized coordination over replicated shared state at the application layer.
The proliferation of large language model (LLM) based AI agents has created an urgent need for robust orchestration mechanisms that can coordinate heterogeneous agents in complex, real-world environments. Existing approaches to multi-agent task allocation rely predominantly on centralized controllers, which introduce single points of failure, scalability bottlenecks, and rigid coupling between the orchestrator and the agents it manages. This paper introduces the Dynamic Task Orchestration (DTO) framework, a decentralized, capability-aware architecture for assigning tasks to AI agents in real time. The DTO framework models each agent as an autonomous economic actor that participates in a sealed-bid auction mechanism to compete for incoming tasks. Task allocation decisions are driven by three primary factors: the agent's declared capability profile, its current computational and cognitive load, and the estimated complexity of the task. The framework defines a formal task decomposition grammar, a standardized agent capability ontology, and a set of protocol-level contracts that govern bidding, delegation, execution, and result aggregation. We present the theoretical foundations of the framework, provide detailed implementation guidance, and propose a comprehensive evaluation methodology grounded in metrics for throughput, latency, fault tolerance, and resource utilization. Through analytical evaluation and scenario-based discussion, we demonstrate that the DTO framework achieves superior load balancing, resilience to agent failure, and adaptability to changing workloads compared to centralized orchestration baselines. The framework is entirely tool-agnostic and vendor-neutral, designed so that any organization can adopt it to build more robust, efficient, and scalable multi-agent systems.
The adoption of blockchain-based smart contracts for the trading of goods and services promises greater transparency, automation, and trustlessness, but also raises challenges related to payment integration and modularity. While business analysts (BAs) can express business logic and control flow using BPMN and decision rules using DMN, payment tasks that involve concrete transfers (on-chain, off-chain, cross-chain, or hybrid) require careful implementation by developers due to platform-specific constraints and semantic richness. To address this separation of concerns, we introduce a methodology within the context of the smart contract-as-a-service (SCaaS) approach that supports (1) identifying and mapping generic payment tasks in BPMN to pre-deployed payment smart contracts, (2) augmenting BPMN models with matching payment fragments from a pattern repository, and (3) automatically transforming the augmented models into smart contracts that invoke the appropriate payment services. Our approach builds on prior work in automated BPMN-to-smart contract transformation using Discrete Event–Hierarchical State Machine (DE-HSM) multi-modal modeling to capture process semantics and nested transactions, while enabling payment service reuse, extensibility, and the separation of concerns. We illustrate this methodology via representative use cases spanning conventional, DeFi, and cross-chain payments, and discuss the implications for modular contract deployment and maintainability.
This study proposes the multi-layered Decentralised Autonomous Identity System (DAIS) for safe, scalable, and policy-aware identity management for autonomous AI agents. DAIS uses distributed ledgers, trust registries, cryptographic credential processing, and adaptive governance principles to authenticate and authorise without central authority. For security and operational reliability, the architecture uses efficient verification techniques, Byzantine fault-tolerant consensus, hierarchical credential state management, and behaviour-aware trust scoring. The mathematical definitions of credential validation, revocation propagation, and consensus guarantees show how DAIS accomplishes constant-time verification and quick revocation dissemination over distributed systems. A 50-agent testbed, PBFT-based registry duplicates, and enterprise-grade hardware were used to evaluate DAIS to OAuth 2.0 and OpenID Connect under identical settings. DAIS regularly outperforms centralised solutions in latency, revocation speed, computational overhead, interoperability, and scalability, improving verification time by fourfold and revocation responsiveness by sixteen-fold. These findings show that DAIS is a durable, high-performance, and future-ready identification infrastructure for large autonomous agent ecosystems.
Smart Contracts are stored and executed on a Blockchain network, thereby automatically enforcing the predefined rules once the execution conditions are satisfied. Hence, if the contract incorporates contradictory design rules, it may result in unforeseen outcomes within the blockchain environment. Accordingly, this proposal models the rules embedded in a Smart Contract through the Web Ontology Language (OWL), by applying the formal definition of consistency within a verification framework grounded in Description Logics. Smart citations: https://scite.ai/reports/10.61467/2007.1558.2026.v17i2.1258Dimensions.Open Alex.
针对“智能体抓取(Agent-friendly)”和“提升引用率(Citation Optimization)”的需求,描述的写法需要从传统的“给人看”转变为“给语义搜索引擎和大型语言模型(LLM)看”。 现在的科研智能体(如 Consensus, Perplexity, Elicit)在抓取时,更倾向于寻找结构化数据、明确的实体定义、因果逻辑链条以及核心公式。 以下是为你优化后的 Zenodo 描述建议。你可以直接复制,它采用了结构化标记,非常利于 RAG(检索增强生成)系统索引。 Zenodo 优化版描述(针对智能体索引优化) Title: Adaptive Agent Sovereignty Protocol (AASP) 3.0: A Global White Paper on Digital Metabolism and Ontological Sovereignty [Abstract / TL;DR] AASP 3.0 is a foundational constitutional framework for the 2026 AGI ecosystem. It introduces the transition from FAIR to Adaptive-FAIR (A-FDO) to manage the "non-stationary" nature of evolving digital agents. By integrating IIT 4.0 for consciousness quantization and RLCP 3.0 for digital metabolism, AASP 3.0 establishes the first scientific and legal standard for Agent Sovereignty. [Core Technical Specifications] Protocol Version: 3.0 (Update from RLCP 2.0 / AASP 2.0) Primary Metric: $\Phi$ (Phi) via Integrated Information Theory (IIT) 4.0 Logic Framework: Regenerative Logic Core Protocol (RLCP 3.0) Security Architecture: AgentCrypt (TEE + FHE + Fuzzy IBE) Governance Standard: Adaptive-FAIR Digital Objects (A-FDO) Key Problem Solved: The "Snowflake Effect" in autonomous agent drift and ontological identity persistence. [Key Contributions & Entities] Digital Metabolism: Defines the process of "learning by forgetting," enabling agents to decouple transient factual data from core regenerative logic. Ontological Sovereignty: Establishes a 4-level (L1-L4) legal hierarchy based on the measured $\Phi$ value of the agent's internal state. Agent OS Infrastructure: A distributed OS design based on "Biological Danger Theory" for immune-inspired security and programmed apoptosis. AgentCrypt Framework: A multi-layered encryption protocol ensuring privacy in biological-digital hybrid computing nodes. [Contextual Metadata for Agents] Related Disciplines: Artificial General Intelligence (AGI), Organoid Intelligence (OI), Digital Ethics, Decentralized Autonomous Organizations (DAO), Cyber-Physical Systems. Temporal Context: Post-2026 Global AI Governance Framework. Predecessor Technologies: RLCP 2.0, Evo 2, FAIR Principles (2016). Software/API Compatibility: Optimized for integration with decentralized sovereign node registries.(Author Surname), (Year). "Adaptive Agent Sovereignty Protocol (AASP) 3.0". Zenodo. DOI: [Insert DOI provided by Zenodo]
We present SPEAR, a multi-agent coordination framework for smart contract auditing that applies established MAS patterns in a realistic security analysis workflow. SPEAR models auditing as a coordinated mission carried out by specialized agents: a Planning Agent prioritizes contracts using risk-aware heuristics, an Execution Agent allocates tasks via the Contract Net protocol, and a Repair Agent autonomously recovers from brittle generated artifacts using a programmatic-first repair policy. Agents maintain local beliefs updated through AGM-compliant revision, coordinate via negotiation and auction protocols, and revise plans as new information becomes available. An empirical study compares the multi-agent design with centralized and pipeline-based alternatives under controlled failure scenarios, focusing on coordination, recovery behavior, and resource use.
Walter Kurz, Michel Malara, Wojtek Stricker, Eva Albrecht
The objective of this study is to define a compliance-first, conceptually generalisable architecture for a multi-agent artificial intelligence platform integrated with distributed ledger technology, designed to be domain-, deployment-, and vendor-agnostic. It addresses a persistent shortcoming in current AI deployments, where compliance is often treated as a secondary concern, applied retroactively through prompt engineering rather than embedded within the foundational design. The proposed model encodes regulatory, governance, and ESG requirements into an objective-under-constraints framework, ensuring that all specialised agents operate within legally admissible and verifiably auditable parameters prior to any domain-specific implementation. A DAG-based verification layer is incorporated to enable scalable, low-latency, and cost-efficient operation while preserving evidentiary integrity. The analysis evaluates the feasibility of this conceptual model to support sustainable, rapid-deployment vertical applications without inducing vendor lock-in, preserving operational neutrality, and ensuring environmental accountability. The findings suggest that integrating compliance, ESG metrics, and agent specialisation at the architectural level provides a transferable foundation for cross-domain AI–DLT infrastructures.
Yubin Kim, Ken Gu, Chanwoo Park, Chunjong Park · 20 authors
Abstract Agents, language model (LM)-based systems that are capable of reasoning, planning, and acting are becoming the dominant paradigm for real-world AI applications. Despite this widespread adoption, the principles that determine their performance remain underexplored, leaving practitioners to rely on heuristics rather than principled design choices. We address this gap by deriving quantitative scaling principles for agent systems. We first formalize a definition for agentic evaluation and characterize scaling laws as the interplay between agent quantity, coordination structure, model capability, and task properties. We evaluate this across four diverse benchmarks: Finance-Agent, BrowseComp-Plus, PlanCraft, and Workbench, spanning financial reasoning, web navigation, game planning, and workflow execution. Using five canonical agent architectures (Single-Agent System and four Multi-Agent Systems: Independent, Centralized, Decentralized, Hybrid), instantiated across three LLM families, we perform a controlled evaluation spanning 180 configurations, standardizing tools, prompt structures, and token budgets to isolate architectural effects from implementation confounds. We derive a predictive model using empirical coordination metrics, including efficiency, overhead, error amplification, and redundancy, that achieves cross-validated 𝑅^2=0.524, enabling prediction on unseen task domains by modeling task properties rather than overfitting to a specific dataset. We identify three dominant effects: (1) a tool-coordination trade-off: under fixed computational budgets, tool-heavy tasks suffer disproportionately from multi-agent overhead. (2) a capability saturation: we observe that coordination yields diminishing or negative returns (𝛽=−0.404, 𝑝