Investigation Report: Yi Wei — Demand-Induced University Patents and Conversion Efficiency (DOI: 10.13653/j.cnki.jqte.20260507.006)

Academic fraud report · Geng Detector

Summary

Verdict: Highly suspicious. The report identifies five anomalies in the paper 'Research on the Impact of Demand-Induced University Patents on Conversion Efficiency — From the Perspective of the Integration of Scientific-Technological Innovation and Industrial Innovation' by Yi Wei, scheduled for issue 5, 2026, in Quantitative and Technical Economic Research. The most serious finding concerns a temporal/factual error: the paper cites the 'Fourth Plenary Session of the 20th CPC Central Committee' in its introduction, although that plenum had not been held as of the publication date, strongly suggesting a fabricated or AI-hallucinated policy citation. Secondary concerns include an unprompted footnote disclaimer stating 'this paper did not use AI,' which is atypical and circumstantial; suspiciously stable regression coefficients across high-dimensional fixed-effects specifications; p-hacking patterns in mechanism analyses; and varying sample sizes across columns. The report rates the overall finding as highly suspicious while explicitly noting that these anomalies are not definitive proof of misconduct and require institutional investigation.

Verdict

Highly suspicious. The principal red flag is an anachronistic policy citation that appears to reference a future event, combined with circumstantial indicators of AI-assisted drafting. Definitive confirmation requires author response, raw data, replication code, and editorial inquiry.

Key findings

Temporal/factual anomaly in the introduction (severity: critical): The paper states that the Fourth Plenary Session of the 20th CPC Central Committee 'clearly pointed out' the need to integrate scientific-technological and industrial innovation. As of May 2026, that plenum had not yet convened (the Third Plenum was held in July 2024). The citation is therefore either fabricated or a large language model hallucination.
Unprompted 'no AI used' disclaimer (severity: high): The title-page footnote declares 'this paper did not use AI.' Such disclaimers are non-standard and, in conjunction with the boilerplate rhetoric and the policy-citation error, raise strong suspicion of AI-assisted generation that the author then publicly denies.
Suspiciously stable regression coefficients (severity: medium): Across columns that add university fixed effects and province-by-year fixed effects, the coefficient on the core explanatory variable (demand-induced patent) remains in a narrow band (0.505 to 0.485). Coefficient stability of this kind under high-dimensional absorption is unusual.
Mechanism-analysis pattern consistent with p-hacking (severity: medium): The novelty and importance interaction terms are insignificant while the 'combinatorial' interaction is 'neatly' positive, mirroring the author's prior theoretical expectation.
Sample-size variation across specifications (severity: medium): In Table 1, N equals 280,334 in columns (1), (2), and (5); 268,990 in column (3); 273,896 in column (4); and drops to 234,289 in column (6) under PPML. The author attributes this to missing values from added controls, but the magnitude of the swings invites scrutiny.
Missing figures and source data (severity: medium): The text references 'Figure 1' repeatedly but the supplied PDF contains no images, precluding pixel-level forensic comparison; raw data and code are not public.

Evidence highlights

Direct quotation, introduction, first paragraph: reference to the 'Fourth Plenary Session of the 20th CPC Central Committee' (二十届四中全会) — event not held as of the reported publication date.
DOI: 10.13653/j.cnki.jqte.20260507.006; journal: 数量经济技术经济研究 (Quantitative and Technical Economic Research), issue 5, 2026.
Table 1 sample-size column values: 280,334; 268,990; 273,896; 234,289.
Core-coefficient range across specifications: 0.505 to 0.485.
Footnote disclaimer on title page: '本文未使用AI'.

Notes

The temporal-citation finding is the strongest single signal; all other items are circumstantial or common in legitimately authored empirical work, although in combination they reinforce suspicion.
Conclusions should be treated as preliminary. The report explicitly acknowledges possible false positives and states that final determinations of misconduct require investigation by qualified institutions.
Recommended follow-up: contact the author for the source of the plenum citation; request public release of data and code to verify coefficient stability and sample construction; raise the issues on PubPeer; notify the journal editorial office regarding any peer-review gaps.