{
  "abstract": "Language models can prepare VASP INCAR files from natural-language requests, but so far only large proprietary cloud models come close to handling the tightly coupled, physics-sensitive settings reliably, a dependence that fits poorly with local, high-throughput materials workflows where privacy, cost, and offline deployment matter. We show that a small language model (SLM) can close this gap. The SLM is fine-tuned on reference VASP calculations and paired with VASPGuard, a deterministic post-processor that checks syntax, workflow, and material-dependent constraints; we call the combined model INCAR-SLM. On INCARBench, a benchmark for VASP INCAR generation, INCAR-SLM built on Qwen3-4B outperforms every general-purpose LLM evaluated, exceeding GPT-5.4 by 15.55 points on the 100-point INCAR Score. Most of this gain comes from fine-tuning, with VASPGuard correcting the errors that remain. We further find that model size matters less than expected: once fine-tuning and post-processing are applied, performance saturates at a few billion parameters, and Qwen3-4B outperforms larger models in the same family.",
  "abstract_shown": true,
  "affiliations": [],
  "article_type": null,
  "arxiv": "2608.05387",
  "author_total": 6,
  "authors": [
    {
      "name": "Xinyue Zhang"
    },
    {
      "name": "Jixiang Li"
    },
    {
      "name": "Bin Shao"
    },
    {
      "name": "Baishun Yang"
    },
    {
      "name": "Zhiyang Liu"
    },
    {
      "name": "Weichao Wang"
    }
  ],
  "bibtex": "https://research.draconitic.com/paper/7n5yzdwtfkfrp.bib",
  "cited_by": 0,
  "doi": "10.48550/arxiv.2608.05387",
  "fulltext_url": null,
  "id": "7n5yzdwtfkfrp",
  "issns": [],
  "landing_url": "https://arxiv.org/abs/2608.05387",
  "language": null,
  "licence": "CC-BY-4.0",
  "licence_label": "CC BY 4.0",
  "licence_source": "article",
  "licence_url": "http://creativecommons.org/licenses/by/4.0/",
  "metadata_licence": "CC0-1.0",
  "native_subjects": [
    [
      "arxiv",
      "cond-mat.mtrl-sci"
    ],
    [
      "arxiv",
      "cs.AI"
    ],
    [
      "openalex",
      "2505|Machine Learning in Materials Science"
    ]
  ],
  "openalex": "W7201869382",
  "pdf_url": "https://arxiv.org/pdf/2608.05387",
  "pmcid": null,
  "pmid": null,
  "published": "2026-08-05",
  "publisher": null,
  "ris": "https://research.draconitic.com/paper/7n5yzdwtfkfrp.ris",
  "sources": [
    {
      "first_seen": "2026-10-30T12:34:04Z",
      "native_id": "2608.05387",
      "source": "arxiv",
      "url": "https://arxiv.org/abs/2608.05387"
    }
  ],
  "subjects": [
    {
      "field": "physics",
      "origin": "arxiv:cond-mat.mtrl-sci",
      "subfield": "condensed-matter"
    }
  ],
  "title": "Fine-Tuning Small Language Models for Reliable VASP INCAR Generation",
  "title_text": "Fine-Tuning Small Language Models for Reliable VASP INCAR Generation",
  "type": "preprint",
  "updated_at": "2026-10-30T12:34:04Z",
  "url": "https://research.draconitic.com/paper/7n5yzdwtfkfrp",
  "venue": "arXiv",
  "versions": []
}