{
  "annotation_id": "6fb72b6f-1be5-4bff-891c-2c682b60c80b",
  "date_created": "2026-02-17T05:53:19.846000Z",
  "date_modified": "2026-02-17T05:53:19.846000Z",
  "file_hash": "f9b70cc31242abff33397bd43e8d2b57a8f270262f2a4d149570f3ab27b24f47",
  "private": false,
  "record": {
    "abstract": "Detecting evasive answers in earnings calls is critical for financial transparency, yet progress is hindered by the lack of large-scale benchmarks. We introduce EvasionBench, comprising 30,000 training samples and 1,000 human-annotated test samples (Cohen\u0027s Kappa 0.835) across three evasion levels. Our key contribution is a multi-model annotation framework leveraging a core insight: disagreement between frontier LLMs signals hard examples most valuable for training. We mine boundary cases where two strong annotators conflict, using a judge to resolve labels. This approach outperforms single-model distillation by 2.4 percent, with judge-resolved samples improving generalization despite higher training loss (0.421 vs 0.393) - evidence that disagreement mining acts as implicit regularization. Our trained model Eva-4B (4B parameters) achieves 81.3 percent accuracy, outperforming its base by 25 percentage points and approaching frontier LLM performance at a fraction of inference cost.",
    "arxiv_id": "2601.09142",
    "authors": [
      "Shijian Ma",
      "Yan Lin",
      "Yi Yang"
    ],
    "categories": [
      "cs.LG",
      "cs.CL"
    ],
    "license": "http://creativecommons.org/licenses/by/4.0/",
    "title": "EvasionBench: Detecting Evasive Answers in Financial Q\u0026A via Multi-Model Consensus and LLM-as-Judge",
    "url": "https://arxiv.org/abs/2601.09142",
    "version": "v1"
  },
  "schema_id": "dorsal/arxiv",
  "source": {
    "execution_id": "4f869e90-6abc-4fdf-ab84-0ece4f6518c5",
    "id": "arXiv Dataset",
    "type": "Model",
    "variant": "snapshot-2026-01-17",
    "version": "0.1.0"
  },
  "user_id": 1000002
}