Eluna: A Graph-Guided Multi-Agent System for Reliable Warehouse SOP Execution
Researchers introduced Eluna, a production-deployed agentic system that encodes Standard Operating Procedures as directed acyclic graphs with progressive disclosure, delegating tasks to parallel sub-agents to enforce procedural compliance under strict time constraints. The system addresses context overload from full SOP specifications that degrades LLM agents.
Entities
Related
GRACE: Graph-Regularized Agentic Context Evolution for Reliable Long-Horizon LLM Agents
A new arXiv paper proposes GRACE, a method that maintains persistent system-level instructions for LLM agents as a typed semantic graph instead of flat text. This graph-regularized approach enables scoped verification and reliable context evolution over long horizons under distribution shift, addressing verification difficulties from accumulated instructions.
GATS framework eliminates LLM calls during agent planning inference
Researchers propose GATS (Graph-Augmented Tree Search), a planning framework that uses a layered world model and UCB1-based tree search to avoid LLM inference during planning, reducing computational cost and stochasticity. The approach outperforms LATS and ReAct on multi-step planning tasks.
Governed multi-agent execution platform with trading engine
Melaya is a platform for designing and executing governed multi-agent workflows, paired with a trading engine. It uses LLMs to orchestrate agents while ensuring security and compliance, targeting operators who need controlled agent execution.
ARCANA: A Reflective Multi-Agent Program Synthesis Framework for ARC-AGI-2 Reasoning
A new multi-agent framework called ARCANA is introduced for solving ARC-AGI-2 tasks under strict test-time and hardware constraints. It decomposes tasks into iterative perception, hypothesis generation, symbolic execution, and reflective refinement using specialized agents communicating through a shared differentiable blackboard.
Local-first coding agent for long autonomous runs
Grinta is a local-first coding agent designed for long autonomous runs. It uses LLMs to autonomously plan and execute coding tasks, solving the problem of needing constant human supervision for extended development sessions.