- Rating
- 3(+1)
- Confidence
- 3
- Informativeness
- 2
- Presentation
- 3
- Soundness
- 2
Read review
Paper summary
This paper investigates positional attention variance (pAV) as a statistical signature for distinguishing AI-generated from natural biological sequences. Using frozen Nucleotide Transformer and ESM2 models, the authors observe differences in attention patterns between natural and generated DNA and proteins. They further develop a DNA screening framework combining sequence statistics and pretrained representations, achieving high detection accuracy through representation fusion and knowledge distillation.
Critical strengths
- Interesting empirical observation: The paper identifies differences in positional attention variance between natural and AI-generated biological sequences. The sequence-shuffling experiments provide evidence that pAV captures information beyond local nucleotide composition.
- Broad evaluation: The authors evaluate multiple DNA generators and protein models, with comparisons against sequence-based baselines and perturbation experiments.
Critical weaknesses
- Limited methodological novelty: The proposed framework primarily relies on existing pretrained models, sequence statistics, and established classification and distillation approaches. While pAV provides an interesting descriptive statistic, its methodological contribution beyond existing sequence-based detection methods remains unclear.
- Limited incremental value and generalization: Simple k-mer entropy and compression features outperform pAV on clean data, and adding pAV does not improve detection performance in the reported experiments. Furthermore, the strongest representation-based detector is evaluated using sequence-level cross-validation, where all generator families appear in both training and evaluation folds. This limits conclusions about generalization to unseen generative models.
Questions for authors
- Generation of synthetic sequences: Were the AI-generated DNA and protein sequences newly generated by the authors or obtained from existing model outputs? Could the authors clarify the generation procedures and sampling settings, and whether these were consistent across models?
- Choice of observer models: What motivated the selection of Nucleotide Transformer for DNA and ESM2 for proteins? Given that pAV varies across observer models, how generalizable is the proposed signature to other biological foundation models?
- Incremental contribution of pAV: Given that simple sequence statistics outperform pAV on clean data, what additional information does pAV provide beyond existing sequence-based features? Could the authors further clarify its contribution to the proposed detection framework?
- Generalization to unseen generators: The representation-based detector is evaluated using sequence-level cross-validation, where all generator families appear in both training and evaluation folds. Could the authors evaluate its performance on entirely unseen generators to better assess generalization?
- Biological interpretation: Beyond differences in attention patterns, is there evidence that pAV captures biologically meaningful sequence properties rather than differences in sequence composition or model-specific attention behavior?
4 comments
Thank you for clarifying method scope and unseen-generator needs (2/2)
2. Performance on Entirely Unseen Generative Models (Weakness 2 and Question 4)
We agree that shared generator families across folds limit any generalization claim. We therefore ran 5-group Leave-One-Family-Out (LOFO), plus three external models absent from every fold (D3LM-from-nt, D3LM-scratch, HyenaDNA-native). External columns are scored against matched out-of-fold naturals (n=2,000) and against 2,500 domain-shifted external naturals. The student row aggregates three seeds.
| Model | LOFO AUROC | LOFO TPR@1.3% | External TPR vs matched | External TPR vs shifted |
|---|---|---|---|---|
CPU-only | 0.9518 | 67.85% | 18.83% | 5.28% |
| NT-2.5B Observer | 0.9689 | 71.73% | 93.83% | n/a |
| Fusion Teacher | 0.9696 | 71.69% | 92.78% | 63.22% |
| Distilled Student | 0.9696 +/- 0.0011 | 77.10 +/- 0.80% | 99.54 +/- 0.16% | 71.59 +/- 1.70% |
Matching external AUROC is 0.9346 (CPU-only), 0.9966 (observer), 0.9965 (teacher), and 0.9992 (student) against matched naturals; against shifted naturals the teacher reaches 0.9832 and the student 0.9877. Crucially, the independent 117M BPE observers also catch D3LM-from-nt (0.9927 and 0.9970 vs matched naturals; 0.9724 and 0.9802 vs shifted naturals). Generalization therefore extends well beyond NT-family models.
3. Synthetic Sequence Provenance and Standardization (Question 1)
Every synthetic corpus in the benchmark panel is sampled from a publicly released checkpoint under its documented default settings. We did not retrain or tune any benchmark generator, which matches real-world external sequence auditing. DNA benchmarks are standardized to 1,000 bp windows. The rebuttal adds three further unseen generators, two of them discrete diffusion models, extending coverage to a different generative paradigm. Our new QC audit also flags low-complexity sampling failures (for example, only 5.2% of GENERator-v2 windows pass the diversity gates). Detection does not depend on those artifacts: after refitting on filtered data only, the observer still leads CPU-only by +19.55 pp at 1.3% FPR.
4. Observer Model Choice and Cross-Architecture Transfer (Question 2)
We selected NT-2.5B and ESM-2 as widely adopted multi-species foundation models with non-overlapping tokenizations. To test whether the signature survives a different architecture and tokenizer, we replicated head discovery on two 117M-parameter BPE models under the same zero-overlap protocol. Macro AUROC is on the four held-out generators, using an unfitted scalar Z(pAV) mean.
| Observer (heads) | Rank transfer (rho) | Scalar Z(pAV) | Random 20-head null |
|---|---|---|---|
NT-2.5B (640) | 0.8473 | 0.8944 | 0.6148 |
DNABERT-S (144) | 0.8594 | 0.9040 | 0.4518 |
DNABERT-2-117M (144) | 0.8837 | 0.9456 | 0.3866 |
The BPE nulls are separated at empirical p = 0.0010 over 1,000 random subsets, establishing clear cross-model generality.
5. Biological Interpretation Beyond Local Composition (Question 5)
We maintain a disciplined boundary and do not map attention heads to specific cis-regulatory motifs. Three converging findings show sensitivity beyond base composition:
- Paired
2-mershuffling holds length, GC, repeats, and exact dinucleotide counts fixed while reducingpAV(median-16.2%on NT-2.5B,p = 6.71e-21; mean-9.13%on DNABERT-2,p = 1.57e-16). - Point substitutions at
5%to10%preservepAVdespite modifying local sequence motifs. - Head selectivity transfers consistently across independent NT and BPE models.
Thank you again for your detailed and balanced evaluation. Your questions directly strengthened our scope clarity and unseen-generator evidence, and we hope these results support raising your score.
Thank you for the detailed rebuttal and substantial additional experiments. I appreciate the authors' efforts to address my concerns. I still have two points that I believe would benefit from clarification:
-
Since adding scalar pAV to the dense representations produces negligible improvements, I think the manuscript should more clearly distinguish the contribution of pAV as a standalone diagnostic statistic from the predictive contribution of the representation-based screening framework.
-
Could the authors clarify how natural and synthetic sequences were matched in the external evaluations, and whether model selection, attention-head selection, and calibration were performed without using any held-out generator information? This would help establish how much of the strong external performance reflects generalizable signatures rather than dataset-specific differences.
But, overall, the rebuttal substantially addresses my main concerns and strengthens the paper. Assuming these evaluation details are clarified, I would be comfortable revising my rating from weak rejection to weak acceptance.
Thank you for the constructive follow-up on pAV scope and external evaluation
We sincerely thank you for reading our rebuttal so carefully and for updating your assessment. Both remaining points are fair, and both improve the paper. Below we give the facts for each, including one new controlled analysis.
1. Separating pAV as a Diagnostic from the Screening Framework
We agree. The evidence supports two distinct contributions, and the revision will present them separately.
| Contribution | Role | Evidence |
|---|---|---|
pAV | Standalone diagnostic: one unfitted, interpretable scalar | Robust to noise (0.867 vs 0.822 for compression at 20% mutation). Head rankings transfer from discovery to held-out generators in three observers (rho 0.8473 to 0.8837). |
| Representation-based screen | Predictive detector: observer, teacher, distilled student | External AUROC 0.9965 (teacher) and 0.9992 (student) vs 0.9346 for CPU-only. Adding scalar pAV changes macro AUROC by less than 0.002. |
2. External Evaluation: Matching, Selection, and Calibration
In our first reply, "matched naturals" meant the in-distribution naturals of the benchmark pool, scored out of fold. Natural and synthetic sequences were aligned on length only (1,000 bp windows). To answer your question directly, we added two controls. First, GC matching: 5% GC bins, 1:1 sampling without replacement in each bin, averaged over 200 resamples (residual GC difference at most 0.003). Second, the two-sided quality gate from our first reply (530/540 external synthetic and 4,496/4,500 natural sequences pass). The pass criterion was fixed before the analysis: after matching, each neural detector must beat CPU-only by at least 0.02 AUROC and stay at or above 0.95 on both comparators.
Macro AUROC over the three unseen external generators (student: mean +/- SD over three seeds):
| Natural comparator | Condition | CPU-only | Teacher | Distilled student |
|---|---|---|---|---|
In-distribution out-of-fold (n=2,000) | As reported | 0.9346 | 0.9965 | 0.9992 +/- 0.0001 |
In-distribution out-of-fold (n=2,000) | GC-matched plus QC | 0.9002 | 0.9952 | 0.9989 +/- 0.0001 |
Domain-shifted external (n=2,500) | As reported | 0.8590 | 0.9832 | 0.9877 +/- 0.0003 |
Domain-shifted external (n=2,500) | GC-matched plus QC | 0.8394 | 0.9746 | 0.9788 +/- 0.0005 |
The criterion is met, and both neural detectors stay above CPU-only for each generator individually. Matching lowers CPU-only (by 0.034 and 0.020) more than the student (by less than 0.001 and by 0.009). Composition differences helped the baseline, not our detectors. A second, independently written implementation reproduces the table (unmatched rows exactly, matched rows within 0.002).
No external-generator information enters any stage of this evaluation:
| Stage | Protocol | Check |
|---|---|---|
| Data | External generators and external naturals appear in no training fold | 0 exact and 0 reverse-complement matches between the 3,040 external sequences and the 5,000-sequence training pool |
| Head selection | None. Observer and teacher read the full 3,200-d representation, with all 640 heads. The student reads raw sequence only | No head subset exists to tune |
| Model selection | Teacher and student keep the configuration stated in the submitted manuscript (student: six epochs, final epoch reported). CPU-only reuses the teacher pipeline | Fixed before the external sets were introduced. No external sequence is used for tuning, early stopping, or checkpoint choice |
| Calibration | Every threshold is set from natural-sequence scores only | Replacing generator scores with arbitrary values in 120 test cases changes no threshold |
AUROC needs no threshold and is our primary external metric. Head selection exists only for the standalone pAV diagnostic, which is not an input to these detectors. There, heads are discovered on a zero-overlap split and tested on four held-out generators, as in our first reply.
Thank you again for your rigorous and constructive follow-up. We commit to the following revisions. The abstract, introduction, and results will state the two contributions separately, and the detection tables will report the with and without pAV ablation. A new external-evaluation protocol subsection will give the matching procedure, the matched AUROC table, and the overlap, selection, and calibration checks. The audit scripts will be released with our code. We hope this fully resolves the remaining points.
Thank you for clarifying method scope and unseen-generator needs (1/2)
We sincerely thank you for acknowledging our sequence-shuffling findings, wide generator coverage, and clear writing. Your questions on novelty, the incremental contribution of
pAVover sequence statistics, observer choice, and unseen generators gave us clear targets. During the rebuttal period, we completed comprehensive cross-observer, quality-control, and unseen-generator studies to answer every point directly.1. Methodological Contribution and Scope Division (Weaknesses 1, 2 and Question 3)
On novelty, prior detection work scores a sequence with external statistics or with a trained classifier head. We instead read the internal attention geometry of a frozen model. Three elements are new. First,
pAVsummarizes how far each query position attends, then measures how that span varies along the sequence. It is a single unfitted scalar, not a learned score. Second, we show the signature is not an artifact of one observer. It replicates on NT-2.5B and on two independent BPE models. Third, we distill the dense observer into a100Mstudent that runs69.09xfaster, which turns the finding into a deployable screen.Our work therefore separates two roles.
pAVis our interpretable structural probe. Dense observer representations, distilled into a compact student, carry the practical screening load. Two measurements fix that division. Under20%point-mutation noise, standalonepAVbeats compression (0.867vs0.822macro AUROC; on Evo-1,0.934vs0.630). Conversely, adding scalarpAVon top of full dense hidden states (3,200-dNT-2.5B or912-dDNABERT) moves macro AUROC by less than0.002in either direction, on the matched pool, on LOFO, and on the clean external generators.Why then are neural representations needed, when simple CPU statistics look strong on unfiltered splits? Because unfiltered pools retain low-complexity synthetic artifacts. Enforcing two-sided quality control (entropy, 6-mer diversity, repeat, and homopolymer gates) leaves
3,211/5,000sequences and changes the ranking. Linear models are refit on filtered data only; our distilled student (seed=42) is rescored and recalibrated on the same rows.1.3%(strict QC)CPU-only(47-d)0.98670.960663.70%0.68630.99470.979683.25%0.72610.99490.980182.60%0.7431pAV0.7516Z(pAV)0.81360.99640.982390.02%Two points follow. The observer gains
+19.55 ppoverCPU-onlyat1.3%FPR once composition artifacts are removed. And on strictly zero-overlap held-out generators under QC,CPU-onlycollapses to0.6863while an unfitted DNABERT-2pAVscalar still reaches0.8136. Under natural domain shift (n=2,500external naturals),CPU-onlyfalls to5.28%TPR at1.3%FPR, while our 3-seed distilled student holds71.59 +/- 1.70%. That student also runs69.09xfaster than the teacher (0.294svs20.317sper500sequences;1.39 GBvs14.36 GBVRAM).