Integrity Report: "An Ultra-Low-Power Transceiver for Bioelectronics with a 30 dB Output Range Digital TX and a -64 dBm Sensitivity RX" (IEEE TCAS-II, DOI 10.1109/TCSII.2026.3737265)

Academic fraud report · Geng Detector

Summary

AI-assisted integrity screening of this IEEE TCAS-II circuit design paper yields a verdict of "highly suspicious" overall, though the aggregated Bayesian synthesis rates the posterior probability of misconduct at 43.1% (95% CI [13.4%, 74%], BF = 14.41), which corresponds to a low-risk aggregate rating. Two substantive findings drive the assessment: (1) the publication footnote states "Received 08 July 2026; revised 04 July 2026; accepted 18 September 2026" — the revision date precedes the receipt date by four days, an internally contradictory timeline that is verbatim in the text (possibly a typesetting error, but not excludable as a record problem); (2) figures explicitly labeled as "measured" (Fig. 9(c) TX output power/efficiency, Fig. 10(c) BER curves) appear unusually smooth, with a cliff-like BER transition between -68 dBm and -64 dBm, raising the possibility of unacknowledged fitting/smoothing or simulation substituted for measurement — though benign curve-fitting practice in circuit papers is a plausible explanation pending raw data. Image forensics clues (copy-move matches at offsets (32,0)/(48,0); noise variance CV=0.83) were reviewed and cleared as page-layout artifacts. Numerical cross-checks of 11 reported values (e.g., 8.2 mW / 6.5 Mbps = 1261.5 pJ/bit; 30.4 dB TX range) were fully consistent. Limitations: AI-generated, not a formal misconduct determination.

Verdict

Highly suspicious (overall screening verdict), with an automated evidence synthesis posterior of 43.1% (95% CI [13.4%, 74%]; BF = 14.41, rated low-risk in aggregate). Two findings warrant author/journal follow-up: a self-contradictory submission timeline and unusually smooth "measured" curves. All image-forensics and numerical-consistency clues were cleared on review. This is not a formal finding of academic misconduct.

Key findings

Finding 1 — Contradictory submission timeline (confirmed, severity: orange): Page 1 footnote reads "Received 08 July 2026; revised 04 July 2026; accepted 18 September 2026." The revision date precedes the receipt date by 4 days. Verbatim in the original text; most likely a proof-stage date-entry error, but submission-record mismanagement cannot be excluded. Contributed 60.2% of the synthesized evidence (logLR = 1.59).
Finding 2 — Suspiciously perfect measured curves (benign explanation possible, severity: yellow): Fig. 9(c) (TX output power and energy efficiency vs. 5-bit codeword) and Fig. 10(c) (BER curves) appear extremely smooth; BER shows a cliff-like drop across -68 dBm to -64 dBm. Text explicitly claims "measured" data ("Fig. 9(c) shows the measured TX output power..."; "the RX sensitivity is improved by up to 4 dB, as shown in Fig. 10(c)"). Real chip measurements typically show unit mismatch, jitter, and channel effects. No smoothing/fitting is declared. Contributed 39.8% of synthesized evidence (logLR = 1.08).
Finding 3 — Image forensics clues cleared: All 8 copy-move matches across img-000–img-009 cluster at only two horizontal offsets — (32,0) (e.g., img-001, 58 pairs) and (48,0) / (32,0) (e.g., img-002, 111 pairs) — consistent across all 10 pages, matching IEEE two-column page-rendering structure rather than localized tampering. Noise variance anomaly (img-004, CV = 0.83) explicable by page background distribution. Not counted toward the suspicious rating.
Finding 4 — Numerical consistency check passed: 8.2 mW ÷ 6.5 Mbps = 1261.5 pJ/bit and 605 µW ÷ 6.5 Mbps = 93.1 pJ/bit match reported energy efficiency; TX dynamic range 3 dBm − (−27.4 dBm) = 30.4 dB matches the claimed 30 dB; Table I FoM values consistent with stated formula and parameters (-64 dBm, 200 µW, DR = 6.5M, TXDR = 30 dB). All 11 cited data points verified in-text. Not counted as evidence.

Evidence highlights

Verbatim footnote on Page 1: "Received 08 July 2026; revised 04 July 2026; accepted 18 September 2026."
DOI: 10.1109/TCSII.2026.3737265 (IEEE Transactions on Circuits and Systems II: Express Briefs, 2026).
21 forensic clues excluded after review as benign, premise-violating, or explanatory.
Synthesis parameters: fixed 5% prior, LR version geng_lr_v3, conditional-independence assumption (may overstate joint strength of correlated evidence).

Notes

Recommended actions: request raw measurement data from authors for Fig. 9(c) and Fig. 10(c) with any smoothing/fitting disclosure; contact the journal editorial office regarding the received/revised date contradiction; institutional-level reporting not currently indicated. AI-generated report for academic discussion only; final misconduct determinations require official institutional investigation, and false positives/negatives are possible.