Geng Integrity Report: "Boosting Semi-Supervised Video Captioning via Learning Candidates Adjusters" (DOI: 10.1145/3652838)

Academic fraud report · Geng Detector

Summary

This Geng-style integrity review flags the paper "Boosting Semi-Supervised Video Captioning via Learning Candidates Adjusters" (Xu et al., ACM TOMM, 2024; DOI: 10.1145/3652838) as highly suspicious (orange). Three principal issues emerge. First, a contradictory publication timeline is documented on the title page: received 5 August 2023, accepted 25 February 2024, yet revised 10 July 2024—a revision date roughly five months after acceptance, which inverts the standard review-order sequence. Second, the affiliation metadata places Beijing Jiaotong University (a Chinese institution) under "Beijing, United States," either suggesting careless template reuse or deliberate international mislabeling. Third, Table 4 shows a non-monotonic trajectory in B@4 (43.66 at T=3 vs. 43.01 at T=4 in the E-step) while ROUGE-L and CIDEr continue to rise, raising concerns of selective reporting. The paper's CV nature precluded Western blot or gel image forensics. Findings reflect text-level evidence only; final adjudication requires publisher investigation.

Verdict

🟠 Highly Suspicious. Multiple metadata inconsistencies and a data irregularity together suggest either significant editorial sloppiness or potential misconduct. The findings are based on text/table inspection only; image forensics was not applicable for this CV paper.

Key findings

Anomalous publication timeline on the paper's copyright footer: Received 5 August 2023; Accepted 25 February 2024; Revised 10 July 2024 — a revision date approximately 4.5 months post-acceptance, which inverts the normal submission → revision → acceptance order.
Geographic mislabeling of affiliation: Beijing Jiaotong University is listed under "Beijing, United States" for authors Zhenjiang Miao and Yigang Cen, despite the institution being located in Beijing, China.
Non-monotonic metric in Table 4 (Page 14): B@4 score drops from 43.66 (T=3, E-step) to 43.01 (T=4, E-step) while ROUGE-L and CIDEr in the same row continue to rise — inconsistent with the expected convergence behavior of an EM-style procedure and suggestive of selective reporting.
Highly templated sentence patterns observed throughout Section 2 (Related Works), e.g., "A novel fully convolutional network (FCN)... is exploited in..."; "A multi-layer sequential model... is proposed in..."; "A syntax-guided hierarchical attention network... is introduced to..."; "A trajectory structured attentional... is presented to..." — reinforcing an overall "assembled" feel.

Evidence highlights

DOI: 10.1145/3652838
Received: 5 August 2023; Accepted: 25 February 2024; Revised: 10 July 2024 (contradiction on title-page metadata).
Affiliation text: "ZHENJIANG MIAO, Beijing Jiaotong University Institute of Information Science, Beijing, United States"; "YIGANG CEN, Beijing Jiaotong University Institute of Information Science, Beijing, United States."
Table 4 values cited: B@4 at T=3 E-step = 43.66; B@4 at T=4 E-step = 43.01 (decrease of 0.65), while ROUGE-L and CIDEr trend upward in the same row.
Image-based PS/pixel-duplication checks were not performed: the paper is a CV algorithm paper whose figures are framework diagrams (Figures 1–3) and visualizations (Figures 4–5), and the source text did not yield high-resolution image data for pixel-level analysis.

Notes

These findings are derived solely from the provided text, metadata, and Table 4; no image-level forensics were applicable.
The timeline anomaly may in principle reflect a journal metadata error rather than author misconduct; independent verification by ACM TOMM editorial office is recommended.
The affiliation mislabeling is most plausibly a LaTeX template carry-over, but cannot be definitively distinguished from intentional misrepresentation without confirmation from the authors or institution.
The non-monotonic B@4 is a single data point and may reflect legitimate run-to-run variance; however, in combination with the other issues it warrants scrutiny of full experimental logs.
This AI-assisted report is for academic discussion only and does not constitute a formal finding of misconduct. Beijing Jiaotong University and the ACM TOMM editorial office should be contacted to clarify the metadata and affiliation inconsistencies.