• Computer Science > Artificial Intelligence [Submitted on 19 Feb 2026] Title:El Agente Gráfico: Structured Execution Graphs for Scientific Agents View PDFAbstract:Large language models (LLMs) are increasingly used to automate scientific workflows, yet their integration with heterogeneous computational tools remains ad hoc and fragile. • Current agentic approaches often rely on unstructured text to manage context and coordinate execution, generating often overwhelming volumes of information that may obscure decision provenance and hinder auditability. • In this work, we present El Agente Gráfico, a single-agent framework that embeds LLM-driven decision-making within a type-safe execution environment and dynamic knowledge graphs for external persistence. • Central to our approach is a structured abstraction of scientific concepts and an object-graph mapper that represents computational state as typed Python objects, stored either in memory or persisted in an external knowledge graph. • This design enables context management through typed symbolic identifiers rather than raw text, thereby ensuring consistency, supporting provenance tracking, and enabling efficient tool orchestration. • We evaluate the system by developing an automated benchmarking framework across a suite of university-level quantum chemistry tasks previously evaluated on a multi-agent system, demonstrating that a single agent, when coupled to a reliable execution engine, can robustly perform complex, multi-step, and paral

Article Summaries:

  • El Agente Gráfico introduces a single‑agent framework that embeds large‑language‑model (LLM) decision‑making inside a type‑safe execution engine and dynamic knowledge graphs. Rather than relying on unstructured text, the system represents computational state as typed Python objects, stored in memory or persisted in an external graph, and uses typed symbolic identifiers for context management. This design preserves provenance, supports auditability, and enables efficient orchestration of heterogeneous scientific tools. The authors benchmark the approach on university‑level quantum‑chemistry tasks, demonstrating robust, multi‑step, parallel execution, and extend the paradigm to conformer ensemble generation and metal‑organic‑framework design. The work shows that abstraction and type safety can scale agentic scientific automation beyond prompt‑centric models.

Sources: