Eluna is a graph-guided, multi-agent framework that encodes SOPs as directed acyclic graphs with progressive disclosure and delegates independent tasks to parallel sub-agents, each with persistent code execution and live data access.
Abstract
Warehouse operations are governed by Standard Operating Procedures (SOPs) that encode complex, multi-system decision logic, which must be executed reliably under strict time constraints, yet LLM agents lack mechanisms to enforce procedural compliance and degrade under the context overload full SOP specifications introduce. We present Eluna, a production-deployed agentic system for reliable SOP execution. Eluna is a graph-guided, multi-agent framework that encodes SOPs as directed acyclic graphs with progressive disclosure and delegates independent tasks to parallel sub-agents, each with persistent code execution and live data access. To meet production latency and accuracy needs, we use asymmetric episodic distillation where a strong teacher is improved through episodic error memories, then a smaller student is fine-tuned on the corrected trajectories with memory stripped, internalizing corrections without inference-time overhead. On a 13-task benchmark and two production applications, our fine-tuned models match or exceed their teacher, beat all larger off-the-shelf baselines, and reach 94% expert agreement on the ticket processing application.
This work introduces an executable benchmark and a budget-aware meta-router that composes heterogeneous operations from raw task text that establishes a reproducible testbed and a bounded proof of concept, not evidence of live-LLM performance.
Natan Vidra, Alina Kapanova, Arun Kanhai et al.· 0 citations
This work presents MetaRoute-Bench, an open, inspectable framework for comparing meta-decision policies under a shared execution model, and releases task generation, policies, traces, tests, and analysis artifacts to support live-system validation.
Natan Vidra, Alina Kapanova, Arun Kanhai et al.· 0 citations
Testing three production engines showed that one achieves 22/22 syntactic TPC-H acceptance but falls short on full semantic validation, function-level pass rates of 94% on non-string categories mask dramatic per- category swings.
HACO is proposed, a runtime control scheme that treats each role request as a reliability-constrained selection problem over candidate agent instances, each coupling a role type, an LLM, and a concrete execution environment.
This work formalizes minimum-sufficient execution and the Agent Cognitive Redundancy Ratio (ACRR), and proposes E3 (Estimate, Execute, Expand): the agent estimates an initial operating point, executes a minimum viable path, and expands scope only when verification fails.
: In complex business scenarios, a customer will often switch quickly among multiple SOPs. The challenge for traditional intelligent customer service systems is that they do not provide the ability to remove siloed information, retain state, and allow for intelligent recovery after interruptions when processing multiple SOPs, requiring users to restart their workflows again from the beginning. In this paper, we will describe an agentic algorithm based on a multi-SOP agentic interruption/resumption perspective. The framework consists of three core capabilities: First, dual-state decoupling separates nondeterministic cognitive decision-making (Coordinator Agent) from deterministic business execution via stateless SOP tools, reducing orchestration complexity and computational overhead. Second, a six-tuple Global State Container serves as the single source of truth for cross-SOP information transfer, enabling automatic reuse of critical business entities and silent propagation of prerequisite parameters across heterogeneous workflows. Third, a Diff Reasoning mechanism measures state discrepancies between frozen snapshots and current global memory, dynamically selecting lossless recovery or safe rollback paths to prevent semantic rollback attacks. The multi-SOP framework reduced the need for users to input the same information more than once, resulting in an 85% reduction to only 12.3, a cross-SOP recovery success rate of 96.3%, and a 73.5% successful chained task completion rate. Average end -to-end time
Zhenggang Wei, Jin-Bin Xu, Zelin Zheng et al.· Journal of Artificial Intell...· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
