{
  "citation": "Acharya, S., Zhang, T. J., Kim, A., et al. (2026). CauSciBench: Can LLMs Automate Causal Inference in Real-World Scientific Research? Proceedings of the 43rd International Conference on Machine Learning (ICML), PMLR 306, 332-384.",
  "deviation_evidence": null,
  "doi": null,
  "excerpt": "Real Gemini-2.5-Flash-Lite Method Sel. Accuracy CoT 51.63",
  "independent": null,
  "item_set": "different",
  "kind": "leaderboard",
  "known_deviation": null,
  "licence": null,
  "locator": "Table 6 (appendix F.1), Dataset Real, Gemini-2.5-Flash-Lite, Method Sel. Accuracy, CoT",
  "metric": "method_selection_accuracy",
  "n_items": 175,
  "page": "26",
  "read_copy_sha256": "d2023d390e33c120c5ddcb6b8d29aae021671877864a21241fc55a407b01e3cc",
  "retrieved_at": "2026-10-15",
  "se": null,
  "se_as_published": null,
  "source_id": "causcibench_real.leaderboard.gemini25flashlite.msa.cot",
  "tolerance_reason": null,
  "units": "%",
  "url": "https://proceedings.mlr.press/v306/acharya26a.html",
  "value": 51.63,
  "value_as_published": "51.63",
  "verified_by": "primary_source",
  "verified_note": "Read from the PMLR PDF (https://raw.githubusercontent.com/mlresearch/v306/main/assets/acharya26a/acharya26a.pdf). Pass@1; prompting strategy CoT; solvers wrote and executed their own code. The paper computes each metric over the queries where it was obtained (its eq. 4-5), over all 175 real queries including matching and glm. The preprint (https://zhijing-jin.com/files/papers/2025_CauSciBench.pdf, sha256 b0d1ea89e177…) prints the same values in its Table 2, the table the econ-bench design cites."
}