{
  "oags": "0.2",
  "status": "alpha",
  "entry": {
    "id": "IB-arxiv-2607.28496-finbert-nonlinear-vs-linear-f1",
    "type": "evidence",
    "claim": "On 41,618 news-stock pairs from the FNSPID dataset, FinBERT sentiment features achieve F1=0.576 under nonlinear models but only F1=0.230 under linear models.",
    "evidence_grade": "E3",
    "state": "active",
    "primary_source_ref": [
      {
        "kind": "arxiv",
        "id": "2607.28496",
        "version": "v1",
        "url": "https://arxiv.org/abs/2607.28496",
        "locator": "abstract"
      }
    ],
    "quantities": {
      "news_stock_pairs": 41618,
      "f1_nonlinear": 0.576,
      "f1_linear": 0.23
    },
    "declared_loss": null,
    "source_title": "Beyond Sentiment: Structured Information Extraction from Financial News",
    "authors": [
      "Daohan Zhu",
      "Sitong Ge",
      "Ruofei Wang",
      "Honggu Chen",
      "Yubo Hou",
      "Tao Wan",
      "Zengchang Qin"
    ],
    "result_origin": "measured",
    "support_quote": "41,618 news--stock pairs from the FNSPID dataset, we find that (i) FinBERT sentiment features exhibit strong predictive power under nonlinear models (F1=0.576) but substantially weaker performance under linear models (F1=0.230)",
    "source_version": "v1",
    "abstract_check": {
      "version": "v1",
      "date": "2026-10-08",
      "quote_verbatim": true
    },
    "domain_v2": "q-fin/ai-finance"
  },
  "expand": [],
  "incoming": [],
  "links": [],
  "declared_losses": [
    "Public evidence-graph projection: internal analytic/theory edges and private provenance are withheld by design.",
    "Alpha: some primary-source refs await precise citation verification."
  ],
  "generated": "2026-10-08"
}