- Rating
- 2
- Confidence
- 3
- Informativeness
- 2
- Presentation
- 2
- Soundness
- 2
Read review
Paper summary
This paper proposes CATS, a conformal-based AI-generated text selection framework that reformulates the conventional binary classification task as a candidate selection problem under false discovery rate (FDR) constraints. Specifically, CATS employs conformal p-values and the Storey–BH procedure to control the expected proportion of falsely flagged human texts within the selected set. Meanwhile, it adaptively adjusts existing detection scores through permutation-equivariant boundary estimation and local selection power optimization, aiming to improve the recall of AI-generated texts under a fixed FDR constraint. The effectiveness of the proposed method is evaluated on 10 data subsets from the MAGE benchmark using 9 different detection scores.
Critical strengths
-
This paper investigates AI-generated text detection from the perspective of FDR control, an evaluation criterion that may have received relatively limited attention in this research area. This provides a novel perspective on the reliability of AI-generated text detection.
-
The paper provides relatively comprehensive theoretical analysis and establishes finite-sample FDR control guarantees under specific exchangeability assumptions. Furthermore, the experiments cover multiple datasets and different types of detection methods, demonstrating that CATS can effectively improve selection power while maintaining FDR control.
Critical weaknesses
-
Motivation and the practical implications of FDR control. Lines 117–127 present a fundamental motivation of this paper: even when the FPR is relatively low, more than half of the flagged texts may still be human-authored, highlighting the necessity of controlling the FDR. However, this argument is based on an imbalanced detection scenario in which human-authored texts substantially outnumber AI-generated texts. If the proportion of AI-generated texts becomes comparable to or exceeds that of human-authored texts, a different concern may arise: even when the FDR appears very low, a substantial number of human-authored texts could still be incorrectly flagged as AI-generated. In other words, a low FDR does not necessarily imply a low FPR, and a high FPR may lead to serious consequences for innocent authors. Therefore, whether FDR control alone sufficiently supports the practical significance emphasized in the first paragraph of the Introduction remains open to discussion.
-
Limited benchmark coverage. The experimental evaluation could be further strengthened by incorporating additional widely used public benchmarks, such as DetectRL, RAID, and RealDet. Evaluations on these datasets would help provide more comprehensive empirical evidence and strengthen the persuasiveness of the study.
-
Selection of baseline methods. The baseline comparisons could be expanded to include more recent and representative detection methods. For statistical-based detectors, commonly adopted methods from the past two years, such as Binoculars and DNA-DetectLLM, could be considered. For supervised detectors, additional comparisons with methods such as DeTeCtive and Biscope would help improve the comprehensiveness of the evaluation.
-
Potential trade-off between FDR and FPR. While the experiments demonstrate improved selection power under controlled FDR, it remains unclear whether the corresponding changes in FPR have been sufficiently examined. Should FPR also be explicitly monitored when optimizing selection power under an FDR constraint? More importantly, does controlling FDR necessarily prevent the FPR from becoming excessively high? Further analysis of the relationship between these two error metrics would help clarify the potential risks of the proposed framework in practical applications.
-
Robustness under text perturbations and varying input lengths. The experimental evaluation could be further extended to consider adversarial text attacks and different input sequence lengths. Additional robustness experiments under these settings would help provide a more comprehensive assessment of the proposed method.
Questions for authors
See Weakness.
6 comments
4. Relationship between the FDR and the FPR [W4]
Thank you for this question. Following the notation in our response to W1, the FPR and the FDP count the same falsely selected human texts but divide them by and , respectively, so that . Since , the FDR guarantee bounds the expected FPR:
which is at most whenever human texts make up at least half of the candidate pool. The reverse does not hold, since a fixed-threshold rule with places no bound on the FDR when machine texts are rare, which is the expected case in high-stakes auditing. Moreover, the auditor can control the FPR at level for any by omitting the Storey estimator in Eq. (7), as shown in our response to W1. We hope this clarifies the relationship between the two error rates.
5. Robustness to adversarial attacks and input length [W5]
Thank you for the suggestion. We add new experiments on adversarial attacks and input length using the same nine detection scores over 500 repetitions. Each table reports the selection power averaged over candidate pools and detection scores (formatted as baseline / our method).
Adversarial attacks. On the candidate pools of RAID [4] and DetectRL [3] in our response to W2, we replace the machine texts with their attacked versions. Each DetectRL row averages the attacks of one category, namely two prompt, three paraphrase, and three perturbation attacks. The results are provided below:
| Dataset | Attack | ||||
|---|---|---|---|---|---|
| RAID | Paraphrase | 0.376 / 0.456 | 0.469 / 0.575 | 0.589 / 0.734 | 0.687 / 0.822 |
| RAID | Synonym | 0.402 / 0.490 | 0.501 / 0.612 | 0.641 / 0.757 | 0.752 / 0.842 |
| RAID | Homoglyph | 0.380 / 0.446 | 0.489 / 0.566 | 0.665 / 0.733 | 0.753 / 0.816 |
| DetectRL | Prompt | 0.332 / 0.380 | 0.399 / 0.467 | 0.476 / 0.576 | 0.537 / 0.658 |
| DetectRL | Paraphrase | 0.241 / 0.283 | 0.301 / 0.366 | 0.375 / 0.473 | 0.431 / 0.555 |
| DetectRL | Perturbation | 0.057 / 0.062 | 0.078 / 0.096 | 0.111 / 0.158 | 0.144 / 0.217 |
Input length. On the varying-length task of DetectRL [3], we group the texts into four candidate pools by word count. The results are provided below:
| Length (words) | ||||
|---|---|---|---|---|
| 0.015 / 0.020 | 0.042 / 0.066 | 0.084 / 0.157 | 0.127 / 0.273 | |
| 100–200 | 0.132 / 0.154 | 0.175 / 0.245 | 0.226 / 0.387 | 0.281 / 0.543 |
| 200–300 | 0.222 / 0.275 | 0.276 / 0.399 | 0.386 / 0.594 | 0.489 / 0.686 |
| 300–360 | 0.406 / 0.459 | 0.494 / 0.616 | 0.593 / 0.757 | 0.652 / 0.813 |
Across all attack types and input lengths, our method consistently improves selection power over the baseline at every target level . Detailed results are reported in Appendix D.5 of the revised version.
References
[1] "Quantifying large language model usage in scientific papers" (Nature Human Behaviour, 2025)
[2] "Delving into LLM-assisted writing in biomedical publications through excess vocabulary" (Science Advances, 2025)
[3] "DetectRL: Benchmarking LLM-Generated Text Detection in Real-World Scenarios" (NeurIPS, 2024)
[4] "RAID: A Shared Benchmark for Robust Evaluation of Machine-Generated Text Detectors" (ACL, 2024)
[5] "Reliably Bounding False Positives: A Zero-Shot Machine-Generated Text Detection Framework via Multiscaled Conformal Prediction" (ACL, 2025)
[6] "Spotting LLMs With Binoculars: Zero-Shot Detection of Machine-Generated Text" (ICML, 2024)
[7] "DNA-DetectLLM: Unveiling AI-Generated Text via a DNA-Inspired Mutation-Repair Paradigm" (NeurIPS, 2025)
[8] "DeTeCtive: Detecting AI-generated Text via Multi-Level Contrastive Learning" (NeurIPS, 2024)
[9] "BiScope: AI-generated Text Detection by Checking Memorization of Preceding Tokens" (NeurIPS, 2024)
Thank you for the comprehensive response and the additional experiments. However, I still have one major concern: while FDR is undoubtedly an important consideration, CATS does not yet seem to have demonstrated its practical utility. In most experiments, the calibration and test sets are tightly coupled in terms of their underlying distributions. Under this setting, many existing calibration methods can also improve performance while providing statistical guarantees.
In real-world detection scenarios, however, the test distribution is generally unknown in advance. How should practitioners determine the distribution of human-written texts used to construct the calibration set? I consider this question central to the practical applicability of the proposed method and would appreciate further discussion.
Thank you for your valuable time and constructive suggestions. We address each concern below.
1. Machine-text proportion in auditing and the FPR under FDR control [W1]
Thank you for this insightful comment. We would like to first clarify the scenario our work targets. We focus on high-stakes auditing, such as manuscript review and coursework auditing, where downstream actions are applied to the flagged texts. Because the use of machine-generated text is restricted and subject to sanctions in these scenarios, machine-generated texts are expected to form a minority of the candidate pool, as supported by recent large-scale studies of scientific writing [1,2]. The imbalanced setting in Lines 117–127 therefore reflects this auditing scenario, where a low FPR can still leave the selected set dominated by human texts and the auditor instead needs to bound the FDR.
Moreover, the FDR guarantee of our method also bounds the FPR. Consider a candidate pool of size m containing m0=π0m human texts. Since 1∨∣S∣≤m in Eq. (4), we have
E[FPR]=E[m0∑j=1m1{j∈S,H0,j is true}]≤E[m0FDP⋅m]=π0FDR≤π0q.
The expected FPR of our method is therefore at most 2q whenever human texts make up at least half of the candidate pool. The reverse does not hold, since a fixed-threshold rule with FPR=α places no bound on the FDR when machine texts are rare, which is the expected case in high-stakes auditing.
When the proportion of machine-generated texts becomes comparable to or exceeds that of human texts, the auditor can control the FPR at level q for any π0 by omitting the Storey estimator in Eq. (7). The selection threshold qkq∗/m then never exceeds q, so every selected text satisfies pj≤q. The super-uniformity of null conformal p-values in Eq. (31) then gives
E[FPR]=m01∑j∈H0P(j∈S)≤m01∑j∈H0P(pj≤q)≤q,
while the FDR remains controlled at π0q≤q. Without the Storey estimator, our method thus controls both the FDR and the FPR at level q, with a more conservative threshold. We hope this discussion addresses your concern.
2. Evaluation on more public benchmarks [W2]
Thank you for the suggestion. We add new experiments on DetectRL [3], RAID [4], and RealDet [5], using the same nine detection scores and settings as the main experiments. The table below reports the selection power over 500 repetitions, averaged over the candidate pools and detection scores of each benchmark (formatted as baseline / our method):
Our method improves selection power on all three benchmarks at every target level. The results for each detection score are reported in Table 15 in Appendix D.3 of the revised version.
3.Additional detection scores [W3]
Thank you for the suggestion. We add four recent detection scores, Binoculars [6], DNA-DetectLLM [7], DeTeCtive [8], and BiScope [9], and evaluate them on RealDet [5]. The table below reports the selection power over 500 repetitions (formatted as baseline / our method):
Our method improves or matches selection power in all settings, and the empirical FDR stays below q in all settings. Detailed results are reported in Appendix D.4 of the revised version.