{
  "abstract": "Lightweight large language models hold promise for scientific applications but remain limited in rule-constrained domains because they can reproduce linguistic patterns without following scientific principles, leading to hallucinations. Here, we show that G-Frame, an adaptive multi-agent framework inspired by Bayesian game and team game principles, enables a closed-loop process for automated data generation, model evaluation, and training. By assigning complementary roles to multiple agents, G-Frame generates structured reasoning data and constructs a specialized corpus containing 363,045 chain-of-thought examples and 199,589 question–answer pairs. Using this corpus together with domain pre-training and adaptive training, the 7-billion-parameter model OmniChem achieves competitive performance across the reported benchmarks and reduces the ChemJudge hallucination rate from 41.95% for the base model to 12.91%. Case studies further illustrate potential applications in molecular design and selected synthesis-planning tasks. These results support G-Frame as an extensible game theory-driven multi-agent framework for improving the reliability of lightweight scientific language models.",
  "abstract_shown": true,
  "affiliations": [],
  "article_type": "research article",
  "arxiv": null,
  "author_total": 10,
  "authors": [
    {
      "name": "Runzhe Liu",
      "orcid": "0009-0006-8879-8089"
    },
    {
      "name": "Biquan Bie"
    },
    {
      "name": "Zihao Wang"
    },
    {
      "name": "Yuchao Ma"
    },
    {
      "name": "Yexin Liu"
    },
    {
      "name": "Xinghai Li"
    },
    {
      "name": "Harry Yang"
    },
    {
      "name": "Wenbo Yang",
      "orcid": "0000-0003-2156-0560"
    },
    {
      "name": "Jinzhe Cao",
      "orcid": "0009-0007-5932-5044"
    },
    {
      "name": "Shengyang Tao",
      "orcid": "0000-0002-0567-8860"
    }
  ],
  "bibtex": "https://research.draconitic.com/paper/6q6x0ae7cv5tn.bib",
  "cited_by": 0,
  "doi": "10.1038/s41467-026-78213-5",
  "fulltext_url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC13736912/",
  "id": "6q6x0ae7cv5tn",
  "issns": [
    "2041-1723"
  ],
  "landing_url": "https://www.nature.com/articles/s41467-026-78213-5",
  "language": "en",
  "licence": "CC-BY-NC-ND-4.0",
  "licence_label": "CC BY-NC-ND 4.0",
  "licence_source": "article",
  "licence_url": "https://creativecommons.org/licenses/by-nc-nd/4.0",
  "metadata_licence": null,
  "native_subjects": [
    [
      "heading",
      "Article"
    ],
    [
      "openalex",
      "1702|Topic Modeling"
    ]
  ],
  "openalex": "W7167903135",
  "pdf_url": "https://www.nature.com/articles/s41467-026-78213-5_reference.pdf",
  "pmcid": "PMC13736912",
  "pmid": "42982833",
  "published": "2026-10-07",
  "publisher": "Springer Science and Business Media LLC",
  "ris": "https://research.draconitic.com/paper/6q6x0ae7cv5tn.ris",
  "sources": [
    {
      "first_seen": "2026-10-30T12:17:38Z",
      "native_id": "10.1038/s41467-026-78213-5",
      "source": "crossref",
      "url": "https://www.nature.com/articles/s41467-026-78213-5"
    },
    {
      "first_seen": "2026-11-09T12:22:15Z",
      "native_id": "PMC13736912",
      "source": "pmc",
      "url": "https://doi.org/10.1038/s41467-026-78213-5"
    }
  ],
  "subjects": [],
  "title": "Game theory driven multi-agent framework mitigates language model hallucination",
  "title_text": "Game theory driven multi-agent framework mitigates language model hallucination",
  "type": "peer-reviewed",
  "updated_at": "2026-11-09T12:22:15Z",
  "url": "https://research.draconitic.com/paper/6q6x0ae7cv5tn",
  "venue": "Nature Communications",
  "versions": []
}