{
  "annotation_id": "6b8201a6-f053-4fc5-8df8-cb36a1fa7509",
  "date_created": "2026-02-17T05:53:24.099000Z",
  "date_modified": "2026-02-17T05:53:24.099000Z",
  "file_hash": "d771e413ca03c3e9f152d7d5cb8cf9833339999e32d26f64a1f57bf69609a048",
  "private": false,
  "record": {
    "abstract": "Scientific and engineering verticals often suffer from data scarcity and strict executability requirements: models must generate not only fluent text, but also syntactically valid, tool-compilable scripts. We present a schema-first alignment framework for building compact, executable domain-specific LLMs in low-resource settings. The framework integrates three core components: (i) large-scale synthetic QA data generation from expert documentation to instill foundational domain knowledge; (ii) a code-centric IR-\u003eDPO workflow that converts verified tool decks into interpretable intermediate representations (IR), performs equivalence-preserving diversification, and constructs preference pairs to directly optimize instruction compliance and code executability; and (iii) a controlled evaluation of Retrieval-Augmented Generation (RAG), showing that while RAG benefits general LLMs, it can marginally degrade the performance of already domain-aligned models.\n  We demonstrate the framework by instantiating TcadGPT for semiconductor Technology Computer-Aided Design (TCAD). Using 1.5M synthetic QA pairs and an IR-driven DPO dataset, TcadGPT attains 85.6% semantic accuracy and an 80.0% syntax pass rate on SDE executability tests, substantially outperforming state-of-the-art general LLMs such as GPT-4o. To probe portability beyond TCAD, we apply the same recipe to the open-source FEM solver Elmer, observing consistent improvements in script-level success rates over general-purpose baselines. All datasets, benchmarks, and code (including P1, P2, and IR-\u003eDPO) are released for reproducibility. Together, these results suggest that the proposed framework provides a robust and reproducible path toward executable LLMs in specialized, data-scarce professional domains.",
    "arxiv_id": "2601.10128",
    "authors": [
      "Di Wang",
      "Zhenhua Wu",
      "Yu Liu",
      "Kai Chang",
      "Shaohua Wu"
    ],
    "categories": [
      "cs.CE",
      "cond-mat.mtrl-sci",
      "cs.SE"
    ],
    "license": "http://creativecommons.org/licenses/by/4.0/",
    "title": "A Generalizable Framework for Building Executable Domain-Specific LLMs under Data Scarcity: Demonstration on Semiconductor TCAD Simulation",
    "url": "https://arxiv.org/abs/2601.10128",
    "version": "v1"
  },
  "schema_id": "dorsal/arxiv",
  "source": {
    "execution_id": "bc86f440-3f41-49cb-b7f3-c13c6dc5159d",
    "id": "arXiv Dataset",
    "type": "Model",
    "variant": "snapshot-2026-01-17",
    "version": "0.1.0"
  },
  "user_id": 1000002
}