Geng Integrity Report — Bi-Objective Heterogeneous Vehicle Routing with Economic Cost and Carbon Emissions under 3D Loading Constraints (DOI: 10.12011/SETP2024-1131)

Academic fraud report · Geng Detector

Summary

Verdict: Highly suspicious. The paper by Ren Zongwei et al., published in Systems Engineering — Theory & Practice (online 2025-06-24), presents a bi-objective VRP model with 3D loading constraints and claims an 'improved genetic algorithm' with 'superior performance'. However, benchmark results on the Augerat dataset (Table 11) reveal solution gaps of roughly 22.12% on A-n32-k5 (957.46 vs BKS 784) and 21.04% on A-n45-k6, far above the 1–5% range typical of modern heuristics. The paper compares mainly against a weaker baseline ('basic GA'), a strawman-style framing that overstates improvements. Additional concerns include a software–hardware timeline inconsistency (PyCharm 2019.3.3, released Nov 2019, paired with Windows 11, released Oct 2021, in a manuscript received May 2024), and reference [28] dated 2025-04-11, post-dating submission. The paper also lacks any statistical significance testing for its comparative claims. Confidence is moderate-to-high for the performance-data issues; environmental/timeline anomalies carry lower confidence pending author clarification.

Verdict

🟠 Highly suspicious. The performance claims on standard benchmarks are inconsistent with the reported numerical results, and secondary issues (environment timeline, reference chronology, missing statistics) further weaken confidence in the experimental section.

Key findings

Strawman benchmark comparison and inflated performance claims (core issue). On Augerat instances, the so-called 'improved GA' yields gaps around 22.12% on A-n32-k5 (957.46 vs. known best 784) and 21.04% on A-n45-k6, with most instances exceeding a 10% gap — well above the 1–5% range typical of modern heuristics. Improvements are demonstrated mainly against a weaker 'basic GA', not against state-of-the-art methods.
Environment timeline inconsistency. Section 5 (p. 15) states the code was developed in PyCharm 2019.3.3 (released November 2019) on Windows 11 (released October 2021), despite the manuscript being received in May 2024.
Chronologically inverted reference. Reference [28] is marked with an access/publication date of 2025-04-11, roughly eleven months after the manuscript receipt date of 2024-05-20.
No statistical significance testing. Comparative statements use terms like 'superior performance' based on single-run averages or best values, with no Wilcoxon, t-test, confidence intervals, or multi-run reporting, despite the stochastic nature of the algorithm.

Evidence highlights

Table 11 (p. 19): Reported gap of ~22.12% on A-n32-k5 (solution 957.46, BKS 784) and ~21.04% on A-n45-k6; majority of test instances show gaps exceeding 10%, some above 30%.
Section 5.3 narrative: Describes 'improved genetic algorithm' achieving 'good results' on Augerat set, contradicting the tabulated numerical gaps.
Section 5 opening (p. 15): Experimental environment stated as Python 3, PyCharm 2019.3.3, Windows 11.
Reference [28] (p. 22): Journal of Chinese Management Science, J/OL entry marked [2025-04-11].
Manuscript metadata: Received 2024-05-20; online publication 2025-06-24; DOI 10.12011/SETP2024-1131.

Notes

Limitations: Only the text of the paper was reviewed; no pixel-level image analysis was performed.
The reference-date inversion could conceivably reflect late-stage editorial additions during typesetting/proof between 2024-05-20 and 2025-06-24, but this should be clarified by the authors.
The PyCharm 2019.3.3 / Windows 11 pairing may reflect outdated local development machines or careless copy-paste, rather than fabricated experiments, but it warrants clarification.
Final determination of misconduct requires an institutional or editorial investigation; this report flags concerns for peer and editorial review.