EVIL-Detect for NLPCC 2026 Shared Task 6: LLM-Generated Text Detection

Hangyu Rong; Hongrui Bao; Yanan Cao; Yubing Ren; Zhuoshang Wang

REVIEW 4 major objections 6 minor 25 references

EVIL-Detect for NLPCC 2026 Shared Task 6: LLM-Generated Text Detection

T0 review · 4 major / 6 minor · reviewed 2026-08-12 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Treating LLM-refined text as a continuous editing extent, plus conflict-aware fusion of four signal types, achieves 0.8888 macro-F1 and first place in NLPCC 2026 Shared Task 6.

desk verdict Competent shared-task system report with a plausible top result, but the testp2 ablation timeline ambiguity undercuts the headline until clarified. read the letter →

arxiv 2608.10698 v1 pith:Z5OF3X6L submitted 2026-08-11 cs.CL

Hongrui Bao , Hangyu Rong , Zhuoshang Wang , Yubing Ren , Yanan Cao This is my paper · ORCID

classification cs.CL

keywords LLM-generatedtextdetectionthree-classclassificationedit-extentregressionconflict-awarefusionzero-shotChineseout-of-distributionrobustnessensemblelearning

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

EVIL-Detect is a system for three-class Chinese text detection, sorting text into human-written (HWT), LLM-generated (LGT), and LLM-refined (HLT) classes. The paper's core move is to treat HLT not as a hard third class but as an intermediate editing state on a continuous axis, with a regression model trained to predict an editing-extent score between a human source and its refined version. At test time the source is not needed: the model predicts the score from the text alone. Around that backbone, the system layers zero-shot likelihood-contrast votes, lexical log-odds statistics, and conservative text rules, then fuses them with two calibrated decision boundaries and a conflict-aware resolver. The paper reports this design is stable under out-of-distribution shifts, achieving macro-F1 of 0.8913 on one hidden test phase and 0.8888 on the other, ranking first in the official evaluation.

What carries the argument

The central object is the editing-extent score $s_E(x)$ produced by an EditLens regression model: a scalar in $[0,1]$ trained to predict $d(h,t)$, the distance between a human source text $h$ and its LLM-refined version $t$, with $r(h)=0$ and $r(g)=1$ as anchors. The distance is computed either by weighted character n-gram cosine distance (Eq. 2) or by semantic phrase-level matching (Eq. 3). This score is discretized with two calibrated boundary pairs $(\tau^{(k)}_1, \tau^{(k)}_2)$ into base labels (Eq. 5); the fusion module keeps the label when the two discretizations agree, and otherwise applies a conflict-aware resolver $R(\cdot)$ that uses Soft-EditLens scores and nine binary LGT-support votes from zero-shot likelihood contrast and lexical log-odds statistics, followed by conservative text rules that flag HTML/XML markup as LGT and rewriting traces as HLT. The machinery's role is to turn an inherently ordinal phenomenon — how much an LLM has edited human text — into a regression target that can be learned without paired sources at test time, and then into deterministic decisions with calibrated thresholds.

What would settle it

Take a sample of held-out (HWT, HLT) pairs from the evaluation distribution, have human annotators rate the degree of editing on a continuous scale, and if the Spearman correlation between those ratings and the EditLens score $s_E(x)$ predicted from the refined text alone is close to zero, the calibrated boundaries in Eq. (5) are not anchored to real editing extent and the fusion is built on a miscalibrated signal.

Watch

Extended reading notes

Core claim

The central discovery is that a three-way HWT/LGT/HLT detector can be made reliable by assigning HLT a soft, instance-dependent editing extent $r(t) = d(h,t)$ anchored at $r(h)=0$ for human text and $r(g)=1$ for fully generated text, rather than treating HLT as a third independent class. The paper instantiates this with two edit-extent regressors — EditLens, using a sweep-selected weighted character n-gram distance (Eq. 2), and Soft-EditLens, using semantic phrase matching and an ordinal bucket head (Eq. 3) — and shows that the best standalone EditLens configuration reaches 0.8494 macro-F1 on the first test phase. The full EVIL-Detect fuses two calibrated discretizations of the EditLens score (Eq. 5) through a conflict-aware resolver that consults Soft-EditLens regression/bucket outputs and a panel of nine binary LGT-support votes from EchoPrompt and lexical statistics, then applies high-precision text rules. On the official test phases the system reports macro-F1 0.8913 (testp1) and 0.8888 (testp2), with HLT remaining the hardest class (F1 around 0.84); the ablation attributes most of the gain over EditLens-only (0.8411) to conflict-aware integration plus rules (0.8816) and the second calibrated boundary (0.8888).

Load-bearing premise

The whole system rests on the premise that the degree to which an LLM has edited a human text can be captured by a scalar distance between the human source and the refined text, and that a model can learn to predict that distance from the text alone, without ever seeing the source at test time.

Editorial extensions

If this is right

For Chinese three-class detection, an edit-extent regression target is a stronger base signal than direct generative classification: direct QLoRA/SFT variants scored 0.1690–0.1896 macro-F1 while EditLens alone scored 0.8494.
The largest remaining error concentration is at the HWT/HLT and HLT/LGT boundaries; conflict-aware fusion cut LGT→HLT errors from 59 to 30 and HLT→LGT errors from 58 to 28 on testp2.
The small gap between testp1 (0.8913) and testp2 (0.8888) indicates that the design transfers across the two out-of-distribution evaluation splits.
Because each auxiliary module (EchoPrompt, lexical statistics) is weak as a standalone three-class predictor (0.63 and 0.55 macro-F1), the system's performance comes from using them as evidence votes in fusion, not as independent classifiers.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

If the editing-extent framing transfers, similar soft-label or ordinal targets could improve binary HWT-vs-AI detection in other languages, since the intermediate class is the same phenomenon everywhere; the paper only demonstrates Chinese news and academic text.
The conflict-aware fusion rules are manually engineered with validation-calibrated thresholds; an implicit testable extension is whether a learned meta-classifier over the same signals would match or beat the hand-written resolver, especially on unseen boundary distributions.
Because the text rules treat raw HTML/XML markup as near-certain LGT, the system would likely misclassify genuinely human-authored markup-heavy content, such as web pages or code documentation, if such content appears in future test sets.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Bottom line: this is a well-structured shared-task system report with a plausible first-place result, but the headline 0.8888 macro-F1 is not independently verifiable and a genuine timeline ambiguity around the testp2 ablation needs to be cleared before I'd trust it as an honest out-of-distribution score.

What is actually new: the ensemble is the contribution. EditLens, EchoPrompt, lexical statistics, and text rules are all prior work, but the editing-extent continuum target (HWT=0, LGT=1, HLT=d(h,t)) and the conflict-aware fusion with validation-calibrated thresholds are a legitimate extension for Chinese three-class OOD detection. Credit goes to the honest component analysis: Table 5 shows most modules are weak standalone, which justifies the multi-signal design rather than hiding behind a single number. The testp2 ablation, if truly post-hoc, shows a real improvement on boundary classes, with LGT→HLT errors cut roughly in half.

Soft spots, in order of severity. First, Section 4.4 says the ablation was run on "released testp2 labels," but the paper never states when those labels became available relative to freezing the fusion thresholds. If thresholds were fitted after seeing testp2, the 0.8888 is a test-set-tuned number and the first-place claim loses its independent value. The paper uses the word "released," which suggests the authors had them, but for how long is unclear. Even if the ablation was purely post-hoc, the ambiguity is trivially resolvable and should be stated explicitly. Second, no error bars, exact threshold values are omitted, and the code link is unversioned, which limits reproducibility. Third, the EditLens distance proxy d(h,t) is validation-fitted and assumes source-free prediction at inference; if the proxy doesn't track human judgments of editing intensity, the entire continuum signal is miscalibrated. That risk is inherent to the design, and the paper only checks it via the rank044 separability sweep, not against human or oracle editing scores.

Who this is for: researchers working on Chinese LLM-text detection, especially three-class and OOD settings, and shared-task participants comparing system architectures. It is a system report, not a conceptual breakthrough.

Recommendation: this deserves peer review, conditionally. A serious referee should ask for the exact timeline of testp2 label release, the calibration threshold values, and a versioned code release. Without that, the central claim stays conditional.

Referee Report

4 major / 6 minor

Summary. The paper presents EVIL-Detect, a multi-signal ensemble system for NLPCC 2026 Shared Task 6, which requires three-class classification of Chinese text into human-written text (HWT), LLM-generated text (LGT), and LLM-refined text (HLT). The system combines an EditLens-inspired edit-extent regression model, a semantic variant called Soft-EditLens, a zero-shot EchoPrompt likelihood-contrast module, lexical frequency statistics, and conservative text rules, all integrated through calibrated decision boundaries and a conflict-aware fusion module. The authors report macro-F1 scores of 0.8913 on testp1 and 0.8888 on testp2, and claim first place in the official evaluation. The paper includes component analysis on testp1 and an ablation study on testp2.

Significance. If the reported results are valid and independent, EVIL-Detect would be a strong published system for Chinese three-class LLM-text detection, with a useful design: using edit-extent regression as the base signal, treating likelihood-based and lexical methods as auxiliary vote sources, and resolving boundary conflicts with a deterministic rule set. The paper's strengths include a released code repository, official leaderboard scores, and explicit component and ablation tables. However, the significance is contingent on two unresolved points: the chronological relationship between the release of testp2 labels and the tuning of fusion thresholds, and the reproducibility of the 'calibrated' thresholds, whose values are not reported. The testp2 ablation, in particular, casts doubt on whether the headline 0.8888 macro-F1 is an independent evaluation or a test-set-tuned number.

major comments (4)

[§4.4, Table 7] The ablation is said to be conducted 'on the released testp2 labels,' but §4.1 describes testp2 as a hidden out-of-distribution test phase. The paper never states whether these labels were available before the fusion thresholds in §3.6 (τ^(k)_1, τ^(k)_2, γ_H, β_T, γ_HL, γ_TL) were finalized. If they were, the reported 0.8888 macro-F1 and the first-place claim are test-set-tuned and not an independent evaluation; if they were not, the text should say so explicitly and should report the validation-tuned configuration's score on testp2 separately. Please clarify the chronological ordering and, if applicable, replace the testp2-based ablation with one on a held-out validation set.
[§3.6] The fusion logic depends on four thresholds and two EditLens boundary settings, but their numerical values and the calibration procedure are not given. The paper says 'validation-calibrated' without defining the validation set, the calibration objective, or whether the calibration was performed once before the test phases. Without this information, the claimed 'calibrated decision boundaries' cannot be reproduced or checked. Provide the exact threshold values, the validation set construction, and the calibration criterion.
[§3.3, Eqs. (1)-(3)] The method assumes that an edit-extent score d(h,t) defined using the paired human source h can be predicted from the target text x alone at inference time. The paper reports downstream macro-F1 but no direct evidence that the predicted scores are calibrated or that the HLT distribution is separable from the two anchors. Please report the correlation between predicted and true edit-extent on a held-out split, and either a calibration curve or histograms of predicted scores by true class, to support the claim that the EditLens score, which anchors the entire fusion in Eq. (5), is reliable.
[Tables 4, 5, 7] All macro-F1 numbers are point estimates from single runs, and testp2 has only 1,152 samples. The reported differences, such as the 4.19-point improvement over EditLens on testp1 and the 0.8411-to-0.8888 improvement in the ablation, may be within sampling noise. Provide bootstrap confidence intervals or a significance test for the main comparisons, especially those that support the robustness claim.

minor comments (6)

[§3.6, Eq. (6)] The variable z_soft is introduced in the resolver description but is not formally defined; define it as the pair (z_r, z_b) or as a named tuple before presenting the conflict-resolution rules.
[Table 3] The nine binary LGT-support votes are summed with equal weight in v(x); state explicitly whether any confidence calibration or per-vote weighting is applied, or whether the votes are intentionally unweighted.
[§4.2] The sentence 'The small gap between the two phases indicates that the multi-signal design remains stable' is stronger than the evidence allows, since there are only two test phases of different sizes; rephrase as 'suggests' or support it with per-phase confidence intervals.
[§4.4] Calling the testp2 analysis an 'ablation' is appropriate only if all configurations were fixed before the testp2 labels were released; otherwise, label it a post-hoc analysis to avoid implying that testp2 was an untouched held-out set.
[§3.3, Eqs. (1)-(3)] Clarify how Eqs. (2) and (3) are combined with the r(h)=0, r(g)=1, r(t)=d(h,t) assignment; in particular, state which distance definition is used for the EditLens rank044 target and how the Soft-EditLens phrasal distance is aligned with the same HWT-to-LGT axis.
[Table 2] The zero-shot module uses both Qwen3.5-4B and Qwen2.5-1.5B instruct/base model pairs, but the table does not specify which checkpoints are instruct models and which are base models; add this information for reproducibility.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: external benchmark anchors the central claim; minor self-citations are not load-bearing.

full rationale

The paper's central derivation is an ensemble whose components are trained or configured on training/validation data and then evaluated on the official hidden test phases. Equation (1) defines the editing-extent target r(t)=d(h,t) using paired HWT/HLT instances, and Eqs. (2)-(3) define the distance d(h,t). The supervised models are trained to predict r(x) from the input text alone, which is a genuine supervised generalization rather than a circular mapping: the test-time input contains no paired source, so the prediction is not an input to the training target. Equations (5)-(6) describe fusion thresholds that are 'validation-calibrated', and the reported macro-F1 values are measured on the official testp1 and testp2 splits; no fitted parameter is renamed as the headline result. Section 4.4's statement that 'We conduct an ablation of the fusion module on the released testp2 labels' does create a timeline ambiguity: if testp2 labels were available before the fusion thresholds were frozen, the 0.8888 testp2 score could have been test-set-tuned. The paper never states the release timing, so this is a verification/contamination risk, not a demonstrated circularity of the derivation chain. The method does cite two works by overlapping authors: EchoPrompt [25] is used as an auxiliary zero-shot vote source, and EnsemJudge [17] is cited as motivation for ensembles. Both are peripheral to the core claim: EditLens [15] is an external prior work, and the final evaluation is against an external shared-task benchmark rather than the authors' own fitted values. Thus no load-bearing step reduces by construction to its own inputs, and the central result is not circular. The score of 1 reflects only the presence of minor self-citations whose contribution is auxiliary and not the basis of the central claim.

Assumptions & free parameters 6 free parameters · 6 assumptions · 0 invented entities

The paper contributes an engineering ensemble; its novelty is the fusion strategy, not a new deductive result. It relies on a substantial number of hand-chosen and validation-fitted parameters and domain assumptions about what signals indicate AI text. None of the fitted parameters are reported numerically, which limits independent assessment.

free parameters (6)

EditLens n-gram target configuration (rank044) = orders 1-7, linearly increasing weights; sweep id 044
Selected via validation-set separability sweep (Sec 3.3); determines the regression targets for EditLens, hence the base signal for fusion.
EditLens decision boundaries tau_1^(k), tau_2^(k) (k=1,2) = not reported numerically
Calibrated on validation to discretize the EditLens score into HWT/HLT/LGT (Eq. 5).
Conflict-aware thresholds gamma_H, beta_T, gamma_HL, gamma_TL = not reported numerically
Calibrated on validation for the conflict resolver R(.) (Sec 3.6).
EchoPrompt model/prompt branches = Qwen2.5-1.5B prefix1; Qwen3.5-4B prefix1; Qwen3.5-4B prefix2
Chosen by hand; outputs feed the LGT vote panel (Table 3).
Lexical statistics smoothing and n-gram retention thresholds = not reported
Determine which n-grams enter the log-odds lexicon and their vote contributions (Sec 3.5).
LoRA hyperparameters = rank 16, alpha 32, dropout 0.05
Chosen for both EditLens and Soft-EditLens; standard but not justified by experiments.

assumptions (6)

domain assumption HLT is an intermediate state on an editing continuum between HWT and LGT, representable by a scalar distance d(h,t).
Defines the regression target r(t)=d(h,t) in Eq. (1); if HLT is not on this continuum, the whole similarity axis is misspecified.
domain assumption Char n-gram overlap or phrase-level soft matching (Eqs. 2-3) is a valid proxy for the degree of AI editing.
Used to construct the EditLens target; no human evaluation of editing extent is reported.
domain assumption EchoPrompt's likelihood-contrast score indicates LGT tendency.
Borrowed from the authors' prior work (Ref [25]); used as zero-shot LGT votes.
domain assumption Lexical n-gram log-odds statistics are class-discriminative after smoothing.
The lexical module only contributes votes; no independent validation beyond the final benchmark.
domain assumption HTML/XML markup and rendering residues are strong evidence of LGT.
High-precision text rules in Sec 3.6; based on observed training examples in Appendix A, not quantified precision.
domain assumption The hidden test sets (testp1, testp2) are truly out-of-distribution and the official labels are trustworthy.
The system's evaluation depends on the shared task's test split and reported leaderboard; the paper does not prove the OOD claim beyond the task description.

how reviews work

0 comments

Cite this review

Pith. "Pith review of EVIL-Detect for NLPCC 2026 Shared Task 6: LLM-Generated Text Detection." pith.science (2026). https://pith.science/paper/Z5OF3X6L

@misc{pith2026260810698,
  author       = {Pith},
  title        = {Pith review of: EVIL-Detect for NLPCC 2026 Shared Task 6: LLM-Generated Text Detection},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/Z5OF3X6L}},
  note         = {Machine review of arXiv:2608.10698}
}

read the original abstract

The rapid development of large language models (LLMs) has increased the need for reliable detection of LLM-generated text, especially in realistic Chinese scenarios involving human-written text (HWT), LLM-generated text (LGT), and LLM-refined text (HLT). This paper presents EVIL-Detect, a multi-signal ensemble framework with conflict-aware fusion for NLPCC 2026 Shared Task 6. The system integrates edit-extent regression, zero-shot likelihood-contrast signals, lexical statistics, and conservative text rules. With calibrated decision boundaries and conflict-aware integration, our system improves robustness under strong out-of-distribution shifts, achieving a macro-F1 score of 0.8888 and ranking first in the official evaluation. Our code is available at https://github.com/bbbbhrrrr/evildetect.

Figures

Figures reproduced from arXiv: 2608.10698 by the authors.

**Figure 1.** Overall architecture of EVIL-Detect. 3.3 Supervised Training Module The supervised training module is inspired by EditLens [15], which formulates AI-edited text detection as continuous edit-extent estimation rather than pure discrete classification. In the original formulation, a similarity metric between a human-written source and its AI-edited version is used as intermediate supervision, and a regression model is… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

25 extracted references · 17 canonical work pages

[1]

arXiv preprint arXiv:2303.08774 (2023)

Achiam, J., Adler, S., Agarwal, S., Ahmad, L., Akkaya, I., Aleman, F.L., et al.: GPT-4 technical report. arXiv preprint arXiv:2303.08774 (2023)

arXiv 2023
[2]

arXiv preprint arXiv:2412.15115 (2024)

Yang, A., Yang, B., Zhang, B., Hui, B., Zheng, B., Yu, B., et al.: Qwen2.5 technical report. arXiv preprint arXiv:2412.15115 (2024)

arXiv 2024
[3]

Computational Linguistics51(1), 275–338 (2025)

Wu, J., Yang, S., Zhan, R., Yuan, Y., Chao, L.S., Wong, D.F.: A survey on LLM- generated text detection: Necessity, methods, and future directions. Computational Linguistics51(1), 275–338 (2025)

work page 2025
[4]

arXiv preprint arXiv:1908.09203 (2019)

Solaiman, I., Brundage, M., Clark, J., Askell, A., Herbert-Voss, A., Wu, J., et al.: Release strategies and the social impacts of language models. arXiv preprint arXiv:1908.09203 (2019)

arXiv 2019
[5]

Advances in Neural Information Processing Systems36, 15077–15095 (2023)

Hu, X., Chen, P.-Y., Ho, T.-Y.: RADAR: Robust AI-text detection via adversarial learning. Advances in Neural Information Processing Systems36, 15077–15095 (2023)

work page 2023
[6]

Advances in Neural Information Processing Systems37, 88320–88347 (2024)

Guo, X., Zhang, S., He, Y., Zhang, T., Feng, W., Huang, H., Ma, C.: DeTeC- tive: Detecting AI-generated text via multi-level contrastive learning. Advances in Neural Information Processing Systems37, 88320–88347 (2024)

work page 2024
[7]

In: Proceed- ings of the 40th International Conference on Machine Learning, pp

Mitchell, E., Lee, Y., Khazatsky, A., Manning, C.D., Finn, C.: DetectGPT: Zero- shot machine-generated text detection using probability curvature. In: Proceed- ings of the 40th International Conference on Machine Learning, pp. 24950–24962. PMLR (2023) 12 H. Bao et al

work page 2023
[8]

In: The Twelfth International Conference on Learning Representations (2024)

Bao, G., Zhao, Y., Teng, Z., Yang, L., Zhang, Y.: Fast-DetectGPT: Efficient zero- shot detection of machine-generated text via conditional probability curvature. In: The Twelfth International Conference on Learning Representations (2024)

work page 2024

Show all 25 references

[9]

In: The Twelfth International Conference on Learning Representations (2024)

Yang, X., Cheng, W., Wu, Y., Petzold, L., Wang, W.Y., Chen, H.: DNA-GPT: Divergent n-gram analysis for training-free detection of GPT-generated text. In: The Twelfth International Conference on Learning Representations (2024)

2024
[10]

In: Proceedings of the 41st International Conference on Machine Learning, pp

Hans, A., Schwarzschild, A., Cherepanova, V., Kazemi, H., Saha, A., Goldblum, M., Geiping, J., Goldstein, T.: Spotting LLMs with binoculars: Zero-shot detection of machine-generated text. In: Proceedings of the 41st International Conference on Machine Learning, pp. 17519–17537...

2024
[11]

In: CCF International Conference on Natural Language Processing and Chinese Computing, pp

Wu, J., Zhan, R., Wang, Q., Yuan, Y., Chao, L.S., Wong, D.F.: Overview of the NLPCC 2025 Shared Task 1: LLM-generated text detection. In: CCF International Conference on Natural Language Processing and Chinese Computing, pp. 263–274. Springer (2025)

2025
[12]

ACM Transactions on Intelligent Systems and Technology17(2), 1–35 (2026)

Tao, Z., Chen, Y., Xi, D., Li, Z., Xu, W.: Toward reliable detection of LLM- generated texts: A comprehensive evaluation framework with CUDRT. ACM Transactions on Intelligent Systems and Technology17(2), 1–35 (2026)

2026
[13]

In: Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pp

Wu, J., Liu, Y., Zhu, C., Zhang, H., Wu, Z., Shi, T., Du, Y., Wang, L., Luo, W., Su, J., Wong, D.F.: DetectRL-X: Towards reliable multilingual and real-world LLM-generated text detection. In: Proceedings of the 64th Annual Meeting of the Association for Computational Linguisti...

2026
[14]

https://github.com/NLP2CT/ NLPCC-2026-Task6-Detection (2026)

NLP2CT Lab: NLPCC 2026 Shared Task 6: The Second Shared Task on LLM-Generated Text Detection. https://github.com/NLP2CT/ NLPCC-2026-Task6-Detection (2026). Accessed 2 July 2026

2026
[15]

arXiv preprint arXiv:2510.03154 (2025)

Thai, K., Emi, B., Masrour, E., Iyyer, M.: EditLens: Quantifying the extent of AI editing in text. arXiv preprint arXiv:2510.03154 (2025)

2025
[16]

Transactions of the Association for Computational Linguistics13, 1812–1831 (2025)

Chen, X., Wu, J., Yang, S., Zhan, R., Wu, Z., Luo, Z., Wang, D., Yang, M., Chao, L.S., Wong, D.F.: RepreGuard: Detecting LLM-generated text by revealing hidden representation patterns. Transactions of the Association for Computational Linguistics13, 1812–1831 (2025)

2025
[17]

In: CCF International Conference on Natural Language Processing and Chinese Computing, Singapore, pp

Wang, Z., Ren, Y., Zhao, G., Zhu, X., Li, H., Cao, Y.: EnsemJudge: Enhancing reli- ability in Chinese LLM-generated text detection through diverse model ensembles. In: CCF International Conference on Natural Language Processing and Chinese Computing, Singapore, pp. 284–295. Sp...

2025
[18]

Advances in Neural Information Processing Systems37, 100369–100401 (2024)

Wu, J., Zhan, R., Wong, D.F., Yang, S., Yang, X., Yuan, Y., Chao, L.S.: DetectRL: Benchmarking LLM-generated text detection in real-world scenarios. Advances in Neural Information Processing Systems37, 100369–100401 (2024)

2024
[19]

arXiv preprint arXiv:2505.09388 (2025)

Yang, A., Li, A., Yang, B., Zhang, B., Hui, B., Zheng, B., Yu, B., et al.: Qwen3 technical report. arXiv preprint arXiv:2505.09388 (2025)

2025 arXiv
[20]

arXiv preprint arXiv:2406.12793 (2024)

Team GLM, Zeng, A., Xu, B., Wang, B., Zhang, C., Yin, D., Zhang, D., et al.: ChatGLM: A family of large language models from GLM-130B to GLM-4 all tools. arXiv preprint arXiv:2406.12793 (2024)

2024 arXiv
[21]

arXiv preprint arXiv:2309.10305 (2023)

Yang, A., Xiao, B., Wang, B., Zhang, B., Bian, C., Yin, C., Lv, C., et al.: Baichuan 2: Open large-scale language models. arXiv preprint arXiv:2309.10305 (2023)

2023 arXiv
[22]

In: International Conference on Learning Representations (2022)

Hu, E.J., Shen, Y., Wallis, P., Allen-Zhu, Z., Li, Y., Wang, S., Wang, L., Chen, W.: LoRA: Low-rank adaptation of large language models. In: International Conference on Learning Representations (2022)

2022
[23]

Advances in Neural Information Processing Systems 36, 10088–10115 (2023) EVIL-Detect for NLPCC 2026 Task 6 13

Dettmers, T., Pagnoni, A., Holtzman, A., Zettlemoyer, L.: QLoRA: Efficient fine- tuning of quantized LLMs. Advances in Neural Information Processing Systems 36, 10088–10115 (2023) EVIL-Detect for NLPCC 2026 Task 6 13

2023
[24]

https://huggingface.co/Qwen (2026)

Qwen Team: Qwen3.5 model cards. https://huggingface.co/Qwen (2026). Accessed 10 July 2026

2026
[25]

arXiv preprint arXiv:2608.05741 (2026) A HTML-like LGT Examples We observed LGT training samples with raw markup or rendering residues

Bao, H., Ren, Y., Cao, Y., You, J., Fang, F., Wang, S.: Once a response, always a response: Detecting LLM-generated text via latent prompt restoration. arXiv preprint arXiv:2608.05741 (2026) A HTML-like LGT Examples We observed LGT training samples with raw markup or rendering...

2026 arXiv

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.

[1] [1]

arXiv preprint arXiv:2303.08774 (2023)

Achiam, J., Adler, S., Agarwal, S., Ahmad, L., Akkaya, I., Aleman, F.L., et al.: GPT-4 technical report. arXiv preprint arXiv:2303.08774 (2023)

arXiv 2023

[2] [2]

arXiv preprint arXiv:2412.15115 (2024)

Yang, A., Yang, B., Zhang, B., Hui, B., Zheng, B., Yu, B., et al.: Qwen2.5 technical report. arXiv preprint arXiv:2412.15115 (2024)

arXiv 2024

[3] [3]

Computational Linguistics51(1), 275–338 (2025)

Wu, J., Yang, S., Zhan, R., Yuan, Y., Chao, L.S., Wong, D.F.: A survey on LLM- generated text detection: Necessity, methods, and future directions. Computational Linguistics51(1), 275–338 (2025)

work page 2025

[4] [4]

arXiv preprint arXiv:1908.09203 (2019)

Solaiman, I., Brundage, M., Clark, J., Askell, A., Herbert-Voss, A., Wu, J., et al.: Release strategies and the social impacts of language models. arXiv preprint arXiv:1908.09203 (2019)

arXiv 2019

[5] [5]

Advances in Neural Information Processing Systems36, 15077–15095 (2023)

Hu, X., Chen, P.-Y., Ho, T.-Y.: RADAR: Robust AI-text detection via adversarial learning. Advances in Neural Information Processing Systems36, 15077–15095 (2023)

work page 2023

[6] [6]

Advances in Neural Information Processing Systems37, 88320–88347 (2024)

Guo, X., Zhang, S., He, Y., Zhang, T., Feng, W., Huang, H., Ma, C.: DeTeC- tive: Detecting AI-generated text via multi-level contrastive learning. Advances in Neural Information Processing Systems37, 88320–88347 (2024)

work page 2024

[7] [7]

In: Proceed- ings of the 40th International Conference on Machine Learning, pp

Mitchell, E., Lee, Y., Khazatsky, A., Manning, C.D., Finn, C.: DetectGPT: Zero- shot machine-generated text detection using probability curvature. In: Proceed- ings of the 40th International Conference on Machine Learning, pp. 24950–24962. PMLR (2023) 12 H. Bao et al

work page 2023

[8] [8]

In: The Twelfth International Conference on Learning Representations (2024)

Bao, G., Zhao, Y., Teng, Z., Yang, L., Zhang, Y.: Fast-DetectGPT: Efficient zero- shot detection of machine-generated text via conditional probability curvature. In: The Twelfth International Conference on Learning Representations (2024)

work page 2024