LatticeMind: A Validation-Gated Agentic Framework for Density Functional Theory Workflows
Abstract
First-principles calculations are rarely single solver calls: they are multistage workflows in which structures, settings, and saved densities must pass correctly between dependent stages. Language models can write a valid single input, but reliability collapses once stages are chained: a dropped dependency, wrong file handoff, or hallucinated keyword can invalidate results even when the solver runs, and such errors surface only after costly execution. We introduce LatticeMind, a validation-gated agentic framework that converts natural-language requests into executable, checkable density functional theory (DFT) projects. A workflow contract specifies required stages, files, and dependencies; deterministic validators inspect the project before any solver launch; failed checks trigger bounded repair; and reports remain traceable to project files. On a 44-case RESCU benchmark spanning total-energy, density-of-states, band-structure, magnetic, phonon workflows, three non-flagship LLM backends reproduced every reference value; 119 of 132 runs passed on the first validated plan and the rest after one repair cycle. Ablations identify the workflow contract and pre-execution validation as the principal reliability mechanisms; a conventional trial-and-repair baseline reaches 71% at twice the token cost per success. The same architecture reproduces every reference value on a 19-case VASP benchmark, demonstrating transfer across solvers. Beyond these task families, LatticeMind generated and validated the 23 inputs of a multistage study of Rh-induced electronic responses at a Rh6/InxGa1−xN photocatalyst interface. These results establish workflow contracts, deterministic validation, bounded repair, and traceable records as a practical foundation for reliable AI-assisted scientific computing, where preventable errors are caught before execution and the solver is never the first validator.
Related
- SQD-Agent: LLM-driven agentic framework for Quantum Chemistry workflows
- Lowering the implementation barrier of neutral-atom quantum computing with agentic workflows
- Evaluating a 4B open-weights local LLM for agentic DFT workflows: a literature reproducibility audit
- Outram Park Part II: Agentic Verification, Validation and Documentation as a Human Review Gate for Nuclear Data and Neutronics Software
- Aitomia: An Agentic Framework for AI-Driven Atomistic and Quantum Chemical Simulations
BibTeX
@misc{vegh2026latticemind,
title = {{LatticeMind: A Validation-Gated Agentic Framework for Density Functional Theory Workflows}},
author = {Nathaniel M. Vegh and Yu Zhu and Raphaël J. Prentki and Félix Beaudoin and Shuaishuai Yuan and Hong Guo},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.26869},
eprint = {2610.26869},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.26869},
}
RIS
TY - UNPB
TI - LatticeMind: A Validation-Gated Agentic Framework for Density Functional Theory Workflows
AU - Nathaniel M. Vegh
AU - Yu Zhu
AU - Raphaël J. Prentki
AU - Félix Beaudoin
AU - Shuaishuai Yuan
AU - Hong Guo
PY - 2026
DA - 2026/10/01
PB - arXiv
DO - 10.48550/arxiv.2610.26869
UR - https://arxiv.org/abs/2610.26869
AB - First-principles calculations are rarely single solver calls: they are multistage workflows in which structures, settings, and saved densities must pass correctly between dependent stages. Language models can write a valid single input, but reliability collapses once stages are chained: a dropped dependency, wrong file handoff, or hallucinated keyword can invalidate results even when the solver runs, and such errors surface only after costly execution. We introduce LatticeMind, a validation-gated agentic framework that converts natural-language requests into executable, checkable density functional theory (DFT) projects. A workflow contract specifies required stages, files, and dependencies; deterministic validators inspect the project before any solver launch; failed checks trigger bounded repair; and reports remain traceable to project files. On a 44-case RESCU benchmark spanning total-energy, density-of-states, band-structure, magnetic, phonon workflows, three non-flagship LLM backends reproduced every reference value; 119 of 132 runs passed on the first validated plan and the rest after one repair cycle. Ablations identify the workflow contract and pre-execution validation as the principal reliability mechanisms; a conventional trial-and-repair baseline reaches 71% at twice the token cost per success. The same architecture reproduces every reference value on a 19-case VASP benchmark, demonstrating transfer across solvers. Beyond these task families, LatticeMind generated and validated the 23 inputs of a multistage study of Rh-induced electronic responses at a Rh_6/In_xGa_1−xN photocatalyst interface. These results establish workflow contracts, deterministic validation, bounded repair, and traceable records as a practical foundation for reliable AI-assisted scientific computing, where preventable errors are caught before execution and the solver is never the first validator.
ER -