Pith. sign in

REVIEW 3 major objections 6 minor 1 cited by

Not all tokens are created equal: Perplexity Attention Weighted Networks for AI generated text detection

T0 review · 3 major / 6 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read PAWN claims that learning to weight token-level likelihood metrics—rather than averaging them—lets a frozen-LLM detector match or beat fine-tuned encoders in-distribution and generalize better to unseen domains and generators.

desk verdict PAWN is a genuine, modest contribution with solid MAGE/M4 evidence and some honest caveats; the RAID split mismatch and missing variance estimates are fixable, so it deserves peer review. read the letter →

arxiv 2501.03940 v3 pith:HTHPON7L submitted 2025-01-07 cs.CL cs.AI

Pablo Miralles-González , Javier Huertas-Tato , Alejandro Martín , David Camacho This is my paper · ORCID
classification cs.CLcs.AI
keywords AI-generatedtextdetectionperplexityweightingnext-tokendistributionmetricsout-of-distributiongeneralizationfrozenLLMdetectormultilingualadversarialrobustness
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

This paper argues that zero-shot AI-text detectors fail partly because they average next-token statistics across all tokens, treating easy and hard predictions alike. It proposes PAWN, a small trained head on top of a frozen LLM that uses the model's hidden states and token positions to weight five next-token distribution metrics before aggregating them. With roughly a million trainable parameters and cached backbone outputs, PAWN reports in-distribution performance competitive with fine-tuned encoder detectors and larger out-of-distribution gains on unseen domains, source models, languages, and decoding settings. The paper's central claim is that semantic-and-position-weighted aggregation is what transfers, not semantic information used directly for classification.

What carries the argument

The central object is the Perplexity Attention Weighted Network (PAWN), a gated aggregation head on a frozen decoder-only LLM. For each token it forms five next-token distribution metrics, processes them with an MLP into F feature channels, and builds G gate logits from the hidden states of the token and its successor plus the normalized position; a softmax across the sequence length turns these gates into weights that sum to one, and the weighted sum of metric features feeds a final classification MLP. The design restricts hidden-state and position information to weighting only, which the paper argues is why PAWN fits in-distribution data without learning domain-specific semantic shortcuts that hurt generalization.

What would settle it

Train PAWN and its mean-aggregation ablation (MPN) on the same corpus, then evaluate both on texts generated with a decoding setting absent from training, such as strong repetition penalty or a very different sampling temperature; if MPN matches or beats PAWN's recall at 5% FPR, the learned weighting is not what drives the reported generalization.

Watch

Extended reading notes

Core claim

On the paper's own terms, the discovery is that a frozen LLM's next-token distributions are a much stronger detection signal than zero-shot use of them suggests, once tokens are weighted by how informative they are. PAWN computes five per-token metrics—log-probability of the occurring token, distribution entropy, maximum log-probability, rank of the occurring token, and top-p coverage—and replaces the usual mean with a weighted sum. The weights come from a small MLP that sees the concatenated hidden states of the current and next token plus a normalized position index, softmaxed across the sequence so they act purely as gates over the metric features. Reported results show PAWN matching or beating the strongest fine-tuned baselines on MAGE's unrestricted in-distribution testbed while suffering smaller performance drops when domain, source model, or language is held out; the same head also retains most of its accuracy under several non-paraphrase adversarial attacks.

Load-bearing premise

The load-bearing premise is that a single frozen LLM's next-token distribution and hidden states contain enough transferable signal to distinguish human from machine text across unseen domains, languages, and generators; if those representations fail to transfer, say to a model with very different tokenization or decoding, the trained head can only reweight the features it already has and cannot recover the missing signal.

Editorial extensions

If this is right

  • Training a detector no longer requires fine-tuning a large language model: a ~1M-parameter head on a frozen backbone reaches the accuracy of fine-tuned encoder detectors in-distribution.
  • Because the backbone is frozen, its hidden states and metrics can be cached once on disk, cutting training time by a factor of 7–10 in the reported setup.
  • Out-of-distribution detection improves in exactly the settings where deployed detectors fail: unseen domains, unseen source models, unseen languages, and unseen decoding strategies such as repetition penalty.
  • The detector's decision boundary is more stable across distribution shifts, so a threshold chosen on training data does not degrade as much when the detector is applied to a new domain.
  • Paraphrasing attacks remain the main failure mode, indicating that robustness to semantic rewriting needs additional machinery such as adversarial training.

Reading between the lines

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

  • If the weighting principle transfers, the same frozen-backbone head could be applied to other per-token signals, such as authorship attribution or machine-translation quality estimation, where token-level confidence is known to be heterogeneous.
  • The paper's own ablation suggests a testable extension: adversarially training the small head against a paraphrasing model could close the paraphrase gap without giving up the generalization advantage, an option the paper names as future work but does not run.
  • The learned per-token weights themselves may serve as an interpretability probe: the appendix already shows that the head upweights sentence-initial discourse markers and downweights word fragments and digits, suggesting the weights encode stylistic priors about which cues betray machine authorship.
  • A cheap stress test of the claim is to swap the backbone for a model with a very different tokenizer and re-run the leave-one-source-model testbeds; if transfer fails there, the approach is tied to the specific backbone's tokenization rather than to a general property of next-token statistics.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 6 minor

Summary. The paper proposes Perplexity Attention Weighted Network (PAWN), a supervised detector that takes a frozen LLM's next-token distribution metrics (log-probability of the next token, entropy, max log-probability, rank, top-p) over a text, and uses a small MLP 'gate' to compute per-token, per-feature weights from the LLM's last hidden states and positional indices. A softmax over sequence length produces weights, and the weighted sum of metric features is passed to a final MLP for binary classification. The head has about 1M-1.6M trainable parameters. Evaluations on MAGE (testbeds 1-8), M4 (domain, language, model leave-one-out), and RAID (decoding strategies, repetition penalty, adversarial attacks) compare PAWN against fine-tuned LMs, zero-shot detectors, and RAID baselines. The central claim is that PAWN is competitive in-distribution and better out-of-distribution than fine-tuned LMs, with more stable decision boundaries and robustness to attacks.

Significance. If supported, the result is practically important: a lightweight head on frozen LLM features yields competitive in-distribution and superior OOD detection at low training cost. The paper's strengths include the use of standard held-out testbeds (MAGE testbeds, M4 leave-one-out), McNemar significance tests on MAGE, a clear ablation isolating the contribution of the weighting mechanism, transparent release of code, and an honest statement of the paraphrasing weakness. The RAID evaluation, however, mixes test splits, and the OOD source-model evidence is confined to one architectural family, so the scope of the headline claims needs tightening.

major comments (3)
  1. [Section 4.5, Tables 10-12] The RAID evaluation mixes results from different test splits and different training sets. PAWN, Longformer, RoBERTa, RADAR-PTM, and fine-tuned RADAR are evaluated on the authors' own split of the RAID open data, whereas the RAID-published baselines (R-B GPT2, R-L GPT2, GLTR, Binoculars, GPTZero, etc.) are evaluated on the hidden test set and trained on different datasets. The footnote to Table 10 acknowledges this, but the text still asserts relative conclusions such as "the methods in the first box ... are generally inferior in absolute terms" and the table presents all rows under one 'Avg.' column. Because split difficulty is not controlled, cross-split comparisons do not support claims of superiority over those baselines. Please either rerun the RAID baselines on the same split used for PAWN, or clearly separate the tables and explicitly state that no cross-split comparison is valid. This issue is load-bearing for the "more robust to adversarial attacks" claim, which relies on Table 12.
  2. [Section 4.1 and 4.4, abstract] The claim that PAWN "generalizes better to unseen domains and source models" is only evidenced for source models from the same broad family as the frozen backbones: all are decoder-only transformer LLMs with subword tokenizers (GPT-2, Llama, Mistral, MPT, Bloomz, ChatGPT, Cohere, Davinci, Dolly, GPT-4). No test includes architectures with fundamentally different tokenization or sequence models (e.g., Mamba, RWKV, char-level models). Because the gating head can only reweight the frozen backbone's next-token metrics, it cannot create signal absent from the backbone; if an unseen model's text has very different surface statistics, the learned softmax weights may be miscalibrated. Appendix B's observation that the highest-weighted tokens are surface/stylistic ("Additionally", "Furthermore", "bacteria") and the near-collapse under paraphrasing (MAGE TB8) reinforce this concern. I recommend either adding at least one out-of-family source model to the OOD evaluation or explicitly restricting the generalization claim to the tested model family.
  3. [Section 4.4, Table 9] In the OOD language tests, PAWN-GPT2 achieves an AUROC of 0.162 on German, far below the random baseline of 0.5. Such a value suggests either label inversion, threshold miscalibration, or an evaluation artifact. The paper should explain this result or correct it, since it undermines confidence in the multilingual evaluation for the GPT-2 backbone (even if the LLaMA backbone performs well).
minor comments (6)
  1. [Section 2.2] "Italia" should be "Italian", and there are several OCR artifacts in the text ("V oight-Kampff", "W A N", "P A WN") that should be cleaned.
  2. [Section 3.1] The entropy metric M^{entropy}_i = \sum_j P_{i,j} \log P_{i,j} is the negative of the usual Shannon entropy; the sign convention should be stated so that the reader knows whether larger values correspond to more random distributions.
  3. [Section 3.1] The definition of M^{rank}_i = rank(\log P_{i,:}, t_{i+1}) / V should specify whether rank is 1-based and whether higher rank corresponds to higher probability.
  4. [Table 10] The header layout (e.g., "Chat No Chat" under "Open Source") is confusing; a clearer presentation would separate open-source and closed-source results into distinct column groups with explicit labels.
  5. [Section 4.3] The ensemble experiments are useful, but the conclusion "we recommend against using ensembled PAWN models" should be justified with a small note on the computational cost of ensembling relative to the observed gains.
  6. [References] References [30] and [31] are pre-prints; please update to the published versions if available.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: PAWN is a supervised head trained on held-out splits; its test-set predictions are not fitted constants and no load-bearing self-citation appears.

full rationale

The paper makes no derivation claim that reduces to its inputs. PAWN extracts five next-token metrics plus hidden states and positions from a frozen LLM, then trains a small gating MLP on MAGE labels; the resulting weights are learned parameters evaluated on MAGE testbeds and on external M4 and RAID splits. No equation defines a predicted quantity in terms of the fitted quantity, and no fitted parameter is renamed as a prediction: the multilingual F1 figures, OOD recalls, and AUROC values are held-out test results. The RAID evaluation uses the authors' own split because the official test set is hidden, and the paper discloses this; that is an evaluation limitation, not circular reasoning. The references are all external benchmark/model papers; no self-citation is load-bearing, and no uniqueness theorem or ansatz is imported from the authors' prior work. Claims about robustness to adversarial attacks are empirical outcomes, not built into the network architecture by construction. Although the OOD generalization is only tested within decoder-only subword-tokenizer families, that is a scope/evidence limitation, not a circularity.

Assumptions & free parameters 5 free parameters · 3 assumptions · 0 invented entities

The central claim depends on the standard detection assumption that frozen-LLM next-token statistics separate human and AI text, on the sufficiency of the chosen five metrics plus hidden-state weighting, and on hand-picked hyperparameters. The paper introduces no new physical or conceptual entities.

free parameters (5)
  • Number of gates G = 256
    Hand-chosen hyperparameter controlling the granularity of token weighting. Listed in Table 1.
  • Number of metric features F = 256
    Hand-chosen output dimension of the metrics MLP. Listed in Table 1.
  • Token dropout ratio = 0.15
    Regularization hyperparameter selected by hand; 15% of tokens are masked during training. Table 1.
  • Max tokens = 512
    Sequence truncation length; affects positional encoding and available context. Table 1.
  • MLP hidden layers and features = 3 layers, 256 hidden features
    Architecture of all MLPs in PAWN, chosen by hand. Table 1.
assumptions (3)
  • domain assumption The next-token distribution of a frozen LLM carries discriminative signal between human and AI text.
    Used throughout Section 3.1 to motivate all five metrics; this is a standard assumption in the cited zero-shot detection literature, but the paper does not prove it.
  • domain assumption The last hidden states and positional indices provide sufficient signal to weight token importance for detection.
    Section 3.1, 'Computing weights', maps hidden states and positions through an MLP to attention weights. No proof is given that this weighting scheme is sufficient or optimal.
  • domain assumption Supervised training on MAGE transfers to unseen domains and models without adaptation.
    The out-of-distribution claims in Sections 4.1 and 4.5 depend on the representativeness of MAGE's training distribution; the paper provides no theoretical justification for transfer.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Not all tokens are created equal: Perplexity Attention Weighted Networks for AI generated text detection." pith.science (2026). https://pith.science/paper/HTHPON7L

@misc{pith2026250103940,
  author       = {Pith},
  title        = {Pith review of: Not all tokens are created equal: Perplexity Attention Weighted Networks for AI generated text detection},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/HTHPON7L}},
  note         = {Machine review of arXiv:2501.03940}
}
read the original abstract

The rapid advancement in large language models (LLMs) has significantly enhanced their ability to generate coherent and contextually relevant text, raising concerns about the misuse of AI-generated content and making it critical to detect it. However, the task remains challenging, particularly in unseen domains or with unfamiliar LLMs. Leveraging LLM next-token distribution outputs offers a theoretically appealing approach for detection, as they encapsulate insights from the models' extensive pre-training on diverse corpora. Despite its promise, zero-shot methods that attempt to operationalize these outputs have met with limited success. We hypothesize that one of the problems is that they use the mean to aggregate next-token distribution metrics across tokens, when some tokens are naturally easier or harder to predict and should be weighted differently. Based on this idea, we propose the Perplexity Attention Weighted Network (PAWN), which uses the last hidden states of the LLM and positions to weight the sum of a series of features based on metrics from the next-token distribution across the sequence length. Although not zero-shot, our method allows us to cache the last hidden states and next-token distribution metrics on disk, greatly reducing the training resource requirements. PAWN shows competitive and even better performance in-distribution than the strongest baselines (fine-tuned LMs) with a fraction of their trainable parameters. Our model also generalizes better to unseen domains and source models, with smaller variability in the decision boundary across distribution shifts. It is also more robust to adversarial attacks, and if the backbone has multilingual capabilities, it presents decent generalization to languages not seen during supervised training, with LLaMA3-1B reaching a mean macro-averaged F1 score of 81.46% in cross-validation with nine languages.

Figures

Figures reproduced from arXiv: 2501.03940 by the authors.

Figure 1
Figure 1. Diagram of the Perplexity Attention Weighted Network. [PITH_FULL_IMAGE:figures/full_fig_p007_1.png] view at source ↗
Figure 2
Figure 2. Average weights given to each token position across 1024 samples, normalized by multiplying with the [PITH_FULL_IMAGE:figures/full_fig_p022_2.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Reasoning Denoiser: Denoising Reasoning Traces for Hallucination Detection in Large Reasoning Models

    cs.AI 2026-07 conditional novelty 6.0 of 10

    REDE removes irrelevant and repetitive reasoning steps using final-answer attention as supervision, improving hallucination detection AUROC by up to ~19 points over unfiltered traces.

Reference graph

Works this paper leans on

38 extracted references · 21 canonical work pages · cited by 1 Pith paper

  1. [1]

    Solaiman et al

    I. Solaiman et al. Release Strategies and the Social Impacts of Language Models. Nov. 13, 2019.DOI: 10.48550/ arXiv.1908.09203 . arXiv: 1908.09203 [cs] . URL: http://arxiv.org/abs/1908.09203 (visited on 01/02/2025). Pre-published

    arXiv doi:10.48550/arxiv.1908.09203 2019

  2. [2]

    MAGE: Machine-Generated Text Detection in the Wild

    Y . Li et al. “MAGE: Machine-Generated Text Detection in the Wild”. In: Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (V olume 1: Long Papers). Bangkok, Thailand: Association for Computational Linguisti...

    doi:10.18653/v1/2024.acl-long.3 2024

  3. [3]

    M4: Multi-Generator, Multi-Domain, and Multi-Lingual Black-Box Machine-Generated Text Detection

    Y . Wang et al. “M4: Multi-Generator, Multi-Domain, and Multi-Lingual Black-Box Machine-Generated Text Detection”. In: Proceedings of the 18th Conference of the European Chapter of the Association for Computational Linguistics (Volume 1: Long Papers). Ed. by Y . Graham and M. Purver. St. Julian’s, Malta: Association for Computational Linguistics, Mar. 202...

    work page 2024

  4. [4]

    RAID: A Shared Benchmark for Robust Evaluation of Machine-Generated Text Detectors

    L. Dugan et al. “RAID: A Shared Benchmark for Robust Evaluation of Machine-Generated Text Detectors”. In: Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (V olume 1: Long Papers). Bangkok, Thailand: Assoc...

    doi:10.18653/v1/2024.acl-long.674 2024

  5. [5]

    Spotting LLMs with Binoculars: Zero-Shot Detection of Machine-Generated Text

    A. Hans et al. “Spotting LLMs with Binoculars: Zero-Shot Detection of Machine-Generated Text”. In: Pro- ceedings of the 41st International Conference on Machine Learning. Ed. by R. Salakhutdinov et al. V ol. 235. Proceedings of Machine Learning Research. PMLR, July 21–27, 2024, pp. 17519–17537. URL: https:// proceedings.mlr.press/v235/hans24a.html

    work page 2024

  6. [6]

    DetectLLM: Leveraging Log Rank Information for Zero-Shot Detection of Machine-Generated Text

    J. Su et al. “DetectLLM: Leveraging Log Rank Information for Zero-Shot Detection of Machine-Generated Text”. In: Findings of the Association for Computational Linguistics: EMNLP 2023. Findings of the Association for Computational Linguistics: EMNLP 2023. Singapore: Association for Computational Linguistics, 2023, pp. 12395–12412. DOI: 10.18653/v1/2023.fin...

    doi:10.18653/v1/2023.findings-emnlp.827 2023

  7. [7]

    DetectGPT: Zero-Shot Machine-Generated Text Detection Using Probability Curvature

    E. Mitchell et al. “DetectGPT: Zero-Shot Machine-Generated Text Detection Using Probability Curvature”. In: Proceedings of the 40th International Conference on Machine Learning . International Conference on 18 Perplexity Attention Weighted Networks for AI generated text detection Machine Learning. PMLR, July 3, 2023, pp. 24950–24962. URL: https://proceedi...

    work page 2023

  8. [8]

    Unsupervised Cross-Lingual Representation Learning at Scale

    A. Conneau et al. “Unsupervised Cross-Lingual Representation Learning at Scale”. In: Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics. Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics. Online: Association for Computational Linguistics, 2020, pp. 8440–

    work page 2020

Show all 38 references
  1. [9]

    A Watermark for Large Language Models

    J. Kirchenbauer et al. “A Watermark for Large Language Models”. In: Proceedings of the 40th International Conference on Machine Learning. International Conference on Machine Learning. PMLR, July 23, 2023

    2023

  2. [10]

    Paraphrasing Evades Detectors of AI-Generated Text, but Retrieval Is an Effective Defense

    K. Krishna et al. “Paraphrasing Evades Detectors of AI-Generated Text, but Retrieval Is an Effective Defense”. In: Thirty-Seventh Conference on Neural Information Processing Systems. 2023. URL: https://openreview. net/forum?id=WbFhFvjjKj

    2023

  3. [11]

    M4GT-Bench: Evaluation Benchmark for Black-Box Machine-Generated Text Detection

    Y . Wang et al. “M4GT-Bench: Evaluation Benchmark for Black-Box Machine-Generated Text Detection”. In: Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). Proceedings of the 62nd Annual Meeting of the Association for...

    2024 doi

  4. [12]

    Overview of the “V oight-Kampff

    J. Bevendorff et al. “Overview of the “V oight-Kampff” Generative AI Authorship Verification Task at PAN and ELOQUENT 2024”. In: Working Notes of CLEF 2024 - Conference and Labs of the Evaluation Forum. Ed. by G. Faggioli et al. CEUR Workshop Proceedings. CEUR-WS.org, Sept. 2024

    2024

  5. [13]

    TURINGBENCH: A Benchmark Environment for Turing Test in the Age of Neural Text Generation

    A. Uchendu et al. “TURINGBENCH: A Benchmark Environment for Turing Test in the Age of Neural Text Generation”. In: Findings of the Association for Computational Linguistics: EMNLP 2021 . Findings of the Association for Computational Linguistics: EMNLP 2021. Punta Cana, Dominic...

    2021 doi

  6. [14]

    Ghostbuster: Detecting Text Ghostwritten by Large Language Models

    V . Verma et al. “Ghostbuster: Detecting Text Ghostwritten by Large Language Models”. In: Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers). Proceedings of the ...

    2024

  7. [15]

    Guo et al

    B. Guo et al. How Close Is ChatGPT to Human Experts? Comparison Corpus, Evaluation, and Detection . Jan. 18, 2023. DOI: 10.48550/arXiv.2301.07597 . arXiv: 2301.07597 [cs] . URL: http://arxiv.org/ abs/2301.07597 (visited on 12/26/2024). Pre-published

    2023 doi arXiv

  8. [16]

    Openai/Gpt-2-Output-Dataset

    OpenAI. Openai/Gpt-2-Output-Dataset. OpenAI, Dec. 24, 2024. URL: https://github.com/openai/gpt- 2-output-dataset (visited on 12/26/2024)

    2024

  9. [17]

    On the Reliability of Watermarks for Large Language Models

    J. Kirchenbauer et al. “On the Reliability of Watermarks for Large Language Models”. In: The Twelfth In- ternational Conference on Learning Representations. 2024. URL: https://openreview.net/forum?id= DEJIDCmWOz

    2024

  10. [18]

    GLTR: Statistical Detection and Visualization of Generated Text

    S. Gehrmann, H. Strobelt, and A. Rush. “GLTR: Statistical Detection and Visualization of Generated Text”. In: Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics: System Demonstrations. Proceedings of the 57th Annual Meeting of the Associati...

    2019 doi

  11. [19]

    Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer

    C. Raffel et al. “Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer”. In:Journal of Machine Learning Research21.140 (2020), pp. 1–67. URL: http://jmlr.org/papers/v21/20-074.html

    2020

  12. [20]

    Fast-DetectGPT: Efficient Zero-Shot Detection of Machine-Generated Text via Conditional Probability Curvature

    G. Bao et al. “Fast-DetectGPT: Efficient Zero-Shot Detection of Machine-Generated Text via Conditional Probability Curvature”. In: The Twelfth International Conference on Learning Representations . 2024. URL: https://openreview.net/forum?id=Bpcgcr8E8Z

    2024

  13. [21]

    BERT: Pre-Training of Deep Bidirectional Transformers for Language Understanding

    J. Devlin et al. “BERT: Pre-Training of Deep Bidirectional Transformers for Language Understanding”. In: Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Pap...

    2019 doi

  14. [22]

    Liu et al.RoBERTa: A Robustly Optimized BERT Pretraining Approach

    Y . Liu et al.RoBERTa: A Robustly Optimized BERT Pretraining Approach. July 26, 2019. arXiv: 1907.11692 [cs]. URL: http://arxiv.org/abs/1907.11692 (visited on 09/01/2023). Pre-published

    2019 arXiv

  15. [23]

    RADAR: Robust AI-Text Detection via Adversarial Learning

    X. Hu, P.-Y . Chen, and T.-Y . Ho. “RADAR: Robust AI-Text Detection via Adversarial Learning”. In:Thirty- Seventh Conference on Neural Information Processing Systems . 2023. URL: https://openreview.net/ forum?id=QGrkbaan79

    2023

  16. [24]

    Does DetectGPT Fully Utilize Perturbation? Bridging Selective Perturbation to Fine-Tuned Contrastive Learning Detector Would Be Better

    S. Liu et al. “Does DetectGPT Fully Utilize Perturbation? Bridging Selective Perturbation to Fine-Tuned Contrastive Learning Detector Would Be Better”. In: Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) . Proceed...

    2024 doi

  17. [25]

    YAKE! Keyword Extraction from Single Documents Using Multiple Local Features

    R. Campos et al. “YAKE! Keyword Extraction from Single Documents Using Multiple Local Features”. In: Information Sciences 509 (Jan. 2020), pp. 257–289. ISSN : 00200255. DOI: 10.1016/j.ins.2019.09.013 . URL: https://linkinghub.elsevier.com/retrieve/pii/S0020025519308588 (visite...

    2020 doi

  18. [26]

    DeTeCtive: Detecting AI-Generated Text via Multi-Level Contrastive Learning

    X. Guo et al. “DeTeCtive: Detecting AI-Generated Text via Multi-Level Contrastive Learning”. In:The Thirty- Eighth Annual Conference on Neural Information Processing Systems . 2024. URL: https://openreview. net/forum?id=cdTTTJfJe3

    2024

  19. [27]

    Fighting Fire with Fire: Can ChatGPT Detect AI-Generated Text?

    A. Bhattacharjee and H. Liu. “Fighting Fire with Fire: Can ChatGPT Detect AI-Generated Text?” In: ACM SIGKDD Explorations Newsletter 25.2 (Mar. 26, 2024), pp. 14–21.ISSN : 1931-0145, 1931-0153. DOI: 10.1145/ 3655103.3655106. URL: https://dl.acm.org/doi/10.1145/3655103.3655106 ...

    2024

  20. [28]

    OUTFOX: LLM-Generated Essay Detection through in-Context Learning with Adversarially Generated Examples

    R. Koike, M. Kaneko, and N. Okazaki. “OUTFOX: LLM-Generated Essay Detection through in-Context Learning with Adversarially Generated Examples”. In: Proceedings of the 38th AAAI Conference on Artificial Intelligence. Vancouver, Canada, Feb. 2024

    2024

  21. [29]

    Li et al

    R. Li et al. Learning to Rewrite: Generalized LLM-Generated Text Detection. Feb. 15, 2025. DOI: 10.48550/ arXiv.2408.04237 . arXiv: 2408.04237 [cs] . URL: http://arxiv.org/abs/2408.04237 (visited on 05/13/2025). Pre-published

    2025 doi arXiv

  22. [30]

    Radford et al

    A. Radford et al. Language Models Are Unsupervised Multitask Learners. OpenAI, 2019

    2019

  23. [31]

    Grattafiori et al

    A. Grattafiori et al. The Llama 3 Herd of Models. Nov. 23, 2024. DOI: 10.48550/arXiv.2407.21783. arXiv: 2407.21783 [cs]. URL: http://arxiv.org/abs/2407.21783 (visited on 12/10/2024). Pre-published

    2024 doi arXiv

  24. [32]

    The Curious Case of Neural Text Degeneration

    A. Holtzman et al. “The Curious Case of Neural Text Degeneration”. In: 8th International Conference on Learning Representations, ICLR 2020, Addis Ababa, Ethiopia, April 26-30, 2020. OpenReview.net, 2020. URL: https://openreview.net/forum?id=rygGQyrFvH

    2020

  25. [33]

    Bag of Tricks for Efficient Text Classification

    A. Joulin et al. “Bag of Tricks for Efficient Text Classification”. In:Proceedings of the 15th Conference of the European Chapter of the Association for Computational Linguistics: Volume 2, Short Papers. Ed. by M. Lapata, P. Blunsom, and A. Koller. Valencia, Spain: Association...

    2017

  26. [34]

    Beltagy, M

    I. Beltagy, M. E. Peters, and A. Cohan. Longformer: The Long-Document Transformer. Dec. 2, 2020. arXiv: 2004.05150 [cs]. URL: http://arxiv.org/abs/2004.05150 (visited on 04/09/2024). Pre-published

    2020 arXiv

  27. [35]

    LLMDet: A Third Party Large Language Models Generated Text Detection Tool

    K. Wu et al. “LLMDet: A Third Party Large Language Models Generated Text Detection Tool”. In:Findings of the Association for Computational Linguistics: EMNLP 2023. Findings of the Association for Computational Linguistics: EMNLP 2023. Singapore: Association for Computational L...

    2023

  28. [1423]

    19 Perplexity Attention Weighted Networks for AI generated text detection

    URL: https://aclanthology.org/N19-1423 (visited on 09/25/2023). 19 Perplexity Attention Weighted Networks for AI generated text detection

    2023

  29. [3019]

    URL: https://www.aclweb.org/anthology/P19-3019 (visited on 12/26/2024)

    2024

  30. [8451]

    URL: https://www.aclweb.org/anthology/2020.acl- main.747 (visited on 12/13/2024)

    DOI: 10.18653/v1/2020.acl-main.747. URL: https://www.aclweb.org/anthology/2020.acl- main.747 (visited on 12/13/2024)

    2020 doi

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.