Pith. sign in

REVIEW 2 major objections 7 minor 38 references

A disentangled MoE with RL routing generalizes AI-text detection to unseen domains.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · deepseek-v4-flash

2026-08-04 00:22 UTC pith:56CQKGNV

load-bearing objection Promising architecture, but the zero-shot gains are inflated by test-set hyperparameter tuning; needs revision before the headline claim can be trusted. the 2 major comments →

arxiv 2511.01192 v2 pith:56CQKGNV submitted 2025-11-03 cs.CL

DEER: Disentangled Mixture of Experts with Instance-Adaptive Routing for Generalizable Machine-Generated Text Detection

Guoxin Ma , Xiaoming Liu , Hongyang Chen , Chengzhengxu Li , Zhaohan Zhang , Shengchao Liu , Yu Lan , Cong Wang
show 1 more author
Chao Shen
This is my paper
classification cs.CL
keywords machine-generated text detectiondomain generalizationmixture of expertsreinforcement learning routingdisentangled representationzero-shot detectionout-of-distribution generalizationLLM text attribution
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The paper argues that existing machine-generated text detectors fail under domain shift because they entangle domain-specific artifacts with universal generation signals, and because expert routing at inference is misaligned with the detection objective. DEER counters this with a two-stage design: domain-supervised training of separate domain-specific and shared experts, then a reinforcement-learned policy that picks which experts to consult for each input based on detection reward, not stylistic similarity. The authors report consistent F1 gains over strong baselines, largest on out-of-domain benchmarks. If right, it offers a modular path to detectors that keep working as new writing domains appear.

Core claim

DEER's central claim is that explicitly decoupling domain-local and domain-invariant knowledge—rather than pooling all training domains into one model—preserves the transferable signatures of machine generation, and that a task-reward-driven router can exploit that structure without domain labels at test time. Concretely, training instantiates one expert per source domain plus a shared expert, with a domain-aware gate; inference uses a policy network to output a distribution over source domains, selects the top-m, and aggregates their logits weighted by selection probability. The reported result is that this beats state-of-the-art detectors on both in-domain and five unseen out-of-domain dat

What carries the argument

The load-bearing mechanism is the two-stage DEER architecture: (1) a Disentangled Mixture-of-Experts with per-domain expert MLPs and a shared expert, trained with domain labels so each expert owns a domain's local patterns while the shared expert learns cross-domain generation signatures; (2) an RL policy network, trained with REINFORCE on detection rewards (accuracy minus a domain-average baseline), that maps each input's encoded state to a soft distribution over source domains. At inference, top-m domain experts plus the shared expert are fused by policy probabilities. The RL reward is what aligns routing with the actual detection objective instead of domain similarity.

Load-bearing premise

The load-bearing premise is that the RL policy trained on source-domain classification rewards will pick experts that generalize to unseen domains, rather than exploiting shortcuts to the binary label (such as always choosing the expert whose bias matches the majority class); the paper does not analyze the policy's action distribution on OOD inputs to rule this out.

What would settle it

A concrete check: take a trained DEER model, feed out-of-domain inputs, record the policy's top-1 domain choice per input, and compare that distribution against a uniform or majority-class baseline. If the policy collapses (e.g., always picks the same domain or the domain whose label prior dominates), and OOD F1 drops to the level of that single expert, the central claim of instance-adaptive routing is falsified. Equivalently, training the same architecture with a randomly initialized policy frozen at uniform weights should match the reported OOD gains if routing were not doing the work.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • Domain-generalizable MGT detection improves without target-domain labels, as shown by average F1 gains of 5.32% out-of-domain and 1.39% in-domain over the strongest baseline.
  • Detectors built this way tolerate input perturbations; DEER keeps higher F1 than metric-based and model-based baselines under repetition, deletion, and replacement attacks.
  • Modular expansion works: adapting to a new domain by adding one expert and fine-tuning only shared experts preserves old-domain performance while cutting trainable parameters by roughly 90% versus full fine-tuning.
  • Because the router is optimized for detection reward rather than style match, aggregating multiple expert groups yields further gains, plateauing near m=3.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • A natural next test is whether the learned policy's selection probabilities correlate with actual domain similarity when labels are available on a held-out domain; if they do not, the RL objective is selecting for something subtler than stylistic match.
  • The reward baseline (average over all domain experts) implies the policy is trained to beat the average expert; a direct implication is that the approach may be most effective when source domains are diverse enough that the average expert is weak, which could be probed by varying domain diversity.
  • The architecture suggests a recipe for other open-set detection tasks where domain labels exist during training but vanish at test time: disentangle per-domain experts, then learn a reward-driven router.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 7 minor

Summary. The paper proposes DEER, a two-stage framework for domain-generalizable machine-generated text (MGT) detection. In the first stage, a disentangled mixture-of-experts (DMoE) module is trained with source-domain supervision: each source domain has dedicated experts, and shared experts capture cross-domain features. In the second stage, the DMoE is frozen and a reinforcement-learning policy is trained to route each input instance to the most useful top-m expert groups at inference, without requiring domain labels. The authors evaluate on five in-domain and five out-of-domain MAGE datasets and report state-of-the-art accuracy and F1, with headline OOD gains of about 5.32% F1 and 3.61% accuracy. Ablations, robustness tests, and an incremental-adaptation experiment are also reported.

Significance. If the out-of-domain result is substantiated, the contribution is meaningful: decoupling domain-local and domain-invariant expertise is a plausible route to more generalizable MGT detection, and the RL-based instance-level router addresses a real deployment gap when domain labels are absent. The paper's Table 2 ablations support the value of both expert types and of the learned routing over oracle/random/classifier alternatives, and the robustness and incremental-adaptation analyses are useful additions. However, the central OOD claim is currently weakened by a selection-protocol problem: key architectural hyperparameters appear to be chosen using the target-domain test sets. The significance of the paper therefore depends on a clean re-evaluation with hyperparameters fixed on source validation data only.

major comments (2)
  1. [Appendix C.1 / Table 1] The headline zero-shot DG-MGT result is compromised by test-set hyperparameter selection. Appendix C.1 states that the numbers of domain-specific experts m1, shared experts m2, and the routing parameter top-m were chosen using the DG-MGT F1 scores shown in Figure 3. Since DG-MGT is the five unseen target domains (CMV, ELI5, WP, TLDR, ROCT), these target-test labels influence architectural choices before the final numbers in Table 1 are produced. The baselines are not given the same target-data privilege. Consequently, the reported average OOD improvements of 5.32% F1 and 3.61% accuracy cannot be interpreted as clean zero-shot generalization; they may partly reflect oracle hyperparameter fitting. The source-domain validation split described in Appendix A.1 should be used for all model selection. Please rerun the comparison with hyperparameters fixed on source validation only, and report b
  2. [§3.2, Eq. (5)-(6), Table 2] The routing policy is trained with rewards derived from the frozen detector's own binary classification outcome on the training instance. This objective rewards selecting any expert that makes the detector's prediction correct, not necessarily the expert that captures domain-relevant structure. The paper does not analyze the learned policy's action distribution, nor does it show that the selected top-m expert groups are sensible on out-of-domain inputs. Please add diagnostics: the policy's entropy and action frequencies on source vs. OOD data, a comparison to a uniform top-m ensemble, and a comparison to a policy trained with an independent domain-classification reward. Without these, the claim that DEER achieves 'task-aligned, domain-agnostic' routing rather than a label shortcut is not directly evidenced.
minor comments (7)
  1. [Abstract / §4.3] The abstract at the beginning of the submission reports average F1 improvements of 1.28% and 2.92%, and accuracy gains of 1.35% and 2.26%, whereas the full-text abstract and Section 4.3 report 1.39% and 5.32% F1 and 1.35% and 3.61% accuracy. These headline numbers must be reconciled.
  2. [§2 Related Work] There are missing citation placeholders in the Mixture-of-Experts paragraph: '(Jacobs et al. 1991;?)' and '(Li et al. 2022; ?;?;?)'. Please complete these references.
  3. [Table 1] Many cells in Table 1 have malformed formatting, e.g., the RoBERTa row reads '88.77 0.64 95.241.07' with no clear separation between accuracy and standard deviation. Please reformat the table so each value is readable.
  4. [Algorithm 1] In Algorithm 1, Step 8 says the predicted label is calculated using Eq. (4), but Eq. (4) is the inference-time expert-group aggregation formula; the training-stage prediction is described by Eq. (3). Please correct the cross-reference.
  5. [Eq. (4)] The notation p(ŷ|d_j, x) is described in the text as a 'prediction logit' but written as a conditional probability. This is confusing: if it is a logit, the expression inside the softmax is a weighted sum of logits; if it is a probability, the operation is not a standard mixture. Please clarify the notation.
  6. [§3.1] The phrase 'enhancing the detector’s DG-MGT detection ceapability' contains a typo ('ceapability' should be 'capability').
  7. [§4.2 / Appendix B] The two domain-generalization baselines MSCL and TACIT are said to be 'adapted to the detection task,' but the adaptation procedure is not specified. Please provide enough detail for reproducibility.

Circularity Check

1 steps flagged

DG-MGT gains in Table 1 are partly fitted to the five 'unseen' target domains via hyperparameter selection (Appendix C.1), so the zero-shot OOD claim is not cleanly out-of-sample.

specific steps
  1. fitted input called prediction [Appendix C.1 (Hyperparameter Analysis), Figure 3; Appendix A.1; §4.1; §4.3/Table 1]
    "we explore various configurations ofm1 andm2 and adopt the best-performing setting as reported in the right heatmap of Figure 3. ... Hyperparameter analysis on DG-MGT.Left: performance variation with different values of top-m expert group selection. Right: F1-score heatmap under different configurations of shared and domain-specific experts. ... DG-MGT measures zero-shot generalization to five unseen target domains (CMV, ELI5, WP, TLDR, and ROCT)."

    The architectural hyperparameters m1, m2, and top-m are selected by maximizing F1 on the five DG-MGT target test domains (CMV, ELI5, WP, TLDR, ROCT). Table 1/§4.3 then reports DEER's zero-shot OOD improvements (e.g., +5.32% F1) on exactly those same domains. The headline OOD number is therefore not a prediction of a fixed model applied to unseen domains; it is partly an optimized fit to the target test distribution. Since no equivalent target-domain hyperparameter tuning is reported for the baselines, the comparison is not cleanly zero-shot and the claimed generalization gain is partially built into the experimental setup.

full rationale

The paper contains no mathematical derivation chain whose outputs equal its inputs, and the RL routing is evaluated against independent external benchmarks, so most of the work is not circular in the derivation sense. The one load-bearing circular/contamination issue is the target-domain hyperparameter selection: Appendix C.1 explicitly states that m1/m2 (and top-m) were chosen by DG-MGT F1, i.e., using the same five 'unseen' domains on which the zero-shot claim is later reported. That makes the central OOD improvement at least partly a fitted value rather than a clean prediction. I also flag two non-circular weaknesses for the record: the Related Work contains malformed citation placeholders ('Jacobs et al. 1991;?'), and the paper does not analyze the learned policy's action distribution to rule out reward hacking; these affect evidence quality but are not circularity. Self-citations to the authors' prior detectors are present but are not load-bearing for the central empirical comparison.

Axiom & Free-Parameter Ledger

4 free parameters · 3 axioms · 0 invented entities

No new physical or theoretical entities are introduced; domain experts, shared experts, and the policy network are implemented as standard MLP components. The main free parameters are architecture sizes and the inference-time expert count, all of which the paper tunes on the DG-MGT test set.

free parameters (4)
  • m1 (number of domain-specific sub-experts per domain) = 5
    Selected from the DG-MGT F1 heatmap in Appendix C.1, i.e., tuned on out-of-domain test data.
  • m2 (number of shared experts) = 6
    Selected from the same DG-MGT F1 heatmap in Appendix C.1.
  • top-m (number of expert groups used at inference) = 3
    Selected by evaluating DG-MGT F1 for m=1..5 in Appendix C.1; performance plateaus after 3.
  • Policy network hidden dimension and optimizer hyperparameters = 768x512; lr=1e-3; weight decay 1e-5
    Hand-chosen, not central to the claim, but needed for exact reproduction.
axioms (3)
  • domain assumption The five chosen source domains (Xsum, HellaSwag, SQuAD, Yelp, Sci) are representative of the MGT-detection domain shift in MAGE.
    Section 4.1 defines the source/target split; the whole DG claim rests on this being a meaningful generalization test.
  • domain assumption The RL policy reward, computed as source-domain classification accuracy of the frozen detector, is a valid proxy for routing quality on unseen domains.
    Section 3.2 defines the reward; no evidence is given that optimizing this proxy transfers to OOD inputs.
  • domain assumption The MAGE benchmark target datasets do not leak into the source training set.
    Section 4 uses five disjoint unseen domains; the result is only a zero-shot claim if this holds.

pith-pipeline@v1.3.0-alltime-deepseek · 15846 in / 13380 out tokens · 143149 ms · 2026-08-04T00:22:59.613847+00:00 · methodology

0 comments
read the original abstract

Detecting machine-generated text has become a critical challenge amid the rapid advancement of LLMs, yet existing detectors degrade severely under domain shift. Through systematic pilot studies, we trace this vulnerability to two fundamental flaws in current generalization strategies, namely the incomplete preservation of domain-specific knowledge during multi-domain training and the misalignment between knowledge retrieval and the detection objective at inference. To address these gaps, we propose DEER, a Disentangled mixturE-of-ExpeRts framework that explicitly decouples domain-local and domain-invariant knowledge into specialized expert modules. Instead of static domain matching, DEER employs a reinforcement learning-driven router that selects expert pathways based on instance-level detection rewards. This task-aligned, domain-agnostic mechanism ensures robust adaptation to unseen distributions by prioritizing detection utility over stylistic resemblance. Extensive experiments demonstrate that DEER consistently outperforms state-of-the-art detectors, achieving average F1 improvements of 1.28% and 2.92%, and accuracy gains of 1.35% and 2.26% on in-domain and out-of-domain datasets, offering reliable generalization for open-world deployment.

Figures

Figures reproduced from arXiv: 2511.01192 by Chao Shen, Chengzhengxu Li, Cong Wang, Guoxin Ma, Hongyang Chen, Shengchao Liu, Xiaoming Liu, Yu Lan, Zhaohan Zhang.

Figure 1
Figure 1. Figure 1: Overview of DEER. The left panel shows the two-stage training process. In the upper part (Sec. 3.1), a disentangled [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Radar plots of pre- and post-incremental per￾formance. (a) Result across source domains before incre￾mental adaptation. (b) Result after adapting to a new unseen domain, evaluated on both source and target domains. Metric RoBERTa-base DEER ↓ Saving (%) Trainable Params (M) 124.6 13.0 89.6% GPU Memory (GB) 14.9 7.6 49.0% Training Time (s) 320 189 40.9% [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Hyperparameter analysis on DG-MGT. Left: performance variation with different values of top-m ex￾pert group selection. Right: F1-score heatmap under differ￾ent configurations of shared and domain-specific experts. C.2 Analysis on Text Length To evaluate the ability of DEER to detect MGTs of vary￾ing lengths, we divide the test samples into segments of 100, 200, 300, and 400 tokens under the WP dataset. As … view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

38 extracted references · 12 linked inside Pith

  1. [1]

    , " * write output.state after.block = add.period write newline

    ENTRY address archivePrefix author booktitle chapter edition editor eid eprint howpublished institution isbn journal key month note number organization pages publisher school series title type volume year label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block FUNCTION init.state.consts #0 'before.a...

  2. [2]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in capitalize " " * FUNCT...

  3. [3]

    AI@Meta. 2024. Llama 3 Model Card

    2024

  4. [4]

    Anthropic. 2024. Claude3. Website

    2024

  5. [5]

    Bao, G.; Zhao, Y.; Teng, Z.; Yang, L.; and Zhang, Y. 2024. Fast-Detect GPT : Efficient Zero-Shot Detection of Machine-Generated Text via Conditional Probability Curvature. In The Twelfth International Conference on Learning Representations

    2024

  6. [6]

    Bhattacharjee, A.; Kumarage, T.; Moraffah, R.; and Liu, H. 2023. Conda: Contrastive domain adaptation for ai-generated text detection. arXiv preprint arXiv:2309.03992

    Pith/arXiv arXiv 2023

  7. [7]

    Bhattacharjee, A.; Moraffah, R.; Garland, J.; and Liu, H. 2024. Eagle: A domain generalization framework for ai-generated text detection. arXiv preprint arXiv:2403.15690

    Pith/arXiv arXiv 2024

  8. [8]

    T.; Li, Z.; Tang, L.; Zhang, L.; et al

    Chen, J.; Zhu, X.; Liu, T.; Chen, Y.; Xinhui, C.; Yuan, Y.; Leong, C. T.; Li, Z.; Tang, L.; Zhang, L.; et al. 2025. Imitate Before Detect: Aligning Machine Stylistic Preference for Machine-Revised Text Detection. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 39, 23559--23567

    2025

  9. [9]

    Dai, Y.; Li, X.; Liu, J.; Tong, Z.; and Duan, L.-Y. 2021. Generalizable person re-identification with relevance-aware mixture of experts. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 16145--16154

    2021

  10. [10]

    Devlin, J.; Chang, M.-W.; Lee, K.; and Toutanova, K. 2019. 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 papers), 4171--4186

    2019

  11. [11]

    Gehrmann, S.; Strobelt, H.; and Rush, A. M. 2019. GLTR: Statistical Detection and Visualization of Generated Text. In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics: System Demonstrations, 111--116

    2019

  12. [12]

    Hans, A.; Schwarzschild, A.; Cherepanova, V.; Kazemi, H.; Saha, A.; Goldblum, M.; Geiping, J.; and Goldstein, T. 2024. Spotting llms with binoculars: Zero-shot detection of machine-generated text. arXiv preprint arXiv:2401.12070

    Pith/arXiv arXiv 2024

  13. [13]

    Hu, X.; Chen, P.-Y.; and Ho, T.-Y. 2023. Radar: Robust ai-text detection via adversarial learning. arXiv preprint arXiv:2307.03838

    Pith/arXiv arXiv 2023

  14. [14]

    A.; Jordan, M

    Jacobs, R. A.; Jordan, M. I.; Nowlan, S. J.; and Hinton, G. E. 1991. Adaptive mixtures of local experts. Neural computation, 3(1): 79--87

    1991

  15. [15]

    Li, B.; Shen, Y.; Yang, J.; Wang, Y.; Ren, J.; Che, T.; Zhang, J.; and Liu, Z. 2022. Sparse mixture-of-experts are domain generalizable learners. arXiv preprint arXiv:2206.04046

    Pith/arXiv arXiv 2022

  16. [16]

    Li, Y.; Li, Q.; Cui, L.; Bi, W.; Wang, Z.; Wang, L.; Yang, L.; Shi, S.; and Zhang, Y. 2024. MAGE : Machine-generated Text Detection in the Wild. In Ku, L.-W.; Martins, A.; and Srikumar, V., eds., Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), 36--53. Bangkok, Thailand: Association for Compu...

    2024

  17. [17]

    Li, Y.; Zhang, Z.; Li, C.; Shen, C.; and Liu, X. 2025. Iron Sharpens Iron: Defending Against Attacks in Machine-Generated Text Detection with Adversarial Training. arXiv:2502.12734

    Pith/arXiv arXiv 2025

  18. [18]

    Liu, S.; Liu, X.; Wang, Y.; Cheng, Z.; Li, C.; Zhang, Z.; Lan, Y.; and Shen, C. 2024. Does detectgpt fully utilize perturbation? bridging selective perturbation to fine-tuned contrastive learning detector would be better. arXiv preprint arXiv:2402.00263

    Pith/arXiv arXiv 2024

  19. [19]

    Liu, X.; and Kong, L. 2024. AI text detection method based on perplexity features with strided sliding window. Working notes of clef

    2024

  20. [20]

    Liu, X.; Zhang, Z.; Wang, Y.; Pu, H.; Lan, Y.; and Shen, C. 2023. CoCo: Coherence-Enhanced Machine-Generated Text Detection Under Low Resource With Contrastive Learning. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, 16167--16188

    2023

  21. [21]

    Liu, Y.; Ott, M.; Goyal, N.; Du, J.; Joshi, M.; Chen, D.; Levy, O.; Lewis, M.; Zettlemoyer, L.; and Stoyanov, V. 2019. Roberta: A robustly optimized bert pretraining approach. arXiv preprint arXiv:1907.11692

    Pith/arXiv arXiv 2019

  22. [22]

    Ma, J.; Zhao, Z.; Yi, X.; Chen, J.; Hong, L.; and Chi, E. H. 2018. Modeling task relationships in multi-task learning with multi-gate mixture-of-experts. In Proceedings of the 24th ACM SIGKDD international conference on knowledge discovery & data mining, 1930--1939

    2018

  23. [23]

    Mireshghallah, F.; Mattern, J.; Gao, S.; Shokri, R.; and Berg-Kirkpatrick, T. 2023. Smaller Language Models are Better Black-box Machine-Generated Text Detectors. arXiv preprint arXiv:2305.09859

    Pith/arXiv arXiv 2023

  24. [24]

    D.; and Finn, C

    Mitchell, E.; Lee, Y.; Khazatsky, A.; Manning, C. D.; and Finn, C. 2023. DetectGPT: Zero-Shot Machine-Generated Text Detection using Probability Curvature. ICML 2023

    2023

  25. [25]

    OpenAI. 2024. GPT-4o System Card

    2024

  26. [26]

    Pudasaini, S.; Miralles, L.; Lillis, D.; and Salvador, M. L. 2025. Benchmarking AI Text Detection: Assessing Detectors Against New Datasets, Evasion Tactics, and Enhanced LLM s. In Alam, F.; Nakov, P.; Habash, N.; Gurevych, I.; Chowdhury, S.; Shelmanov, A.; Wang, Y.; Artemova, E.; Kutlu, M.; and Mikros, G., eds., Proceedings of the 1stWorkshop on GenAI Co...

    2025

  27. [27]

    Qu, J.; Faney, T.; Wang, Z.; Gallinari, P.; Yousef, S.; and de Hemptinne, J.-C. 2022. Hmoe: Hypernetwork-based mixture of experts for domain generalization. arXiv preprint arXiv:2211.08253

    Pith/arXiv arXiv 2022

  28. [28]

    Ren, X.; Zhou, P.; Meng, X.; Huang, X.; Wang, Y.; Wang, W.; Li, P.; Zhang, X.; Podolskiy, A.; Arshinov, G.; et al. 2023. Pangu- \ Sigma \ : Towards trillion parameter language model with sparse heterogeneous computing. arXiv preprint arXiv:2303.10845

    Pith/arXiv arXiv 2023

  29. [29]

    Shum, K.; Diao, S.; and Zhang, T. 2023. Automatic Prompt Augmentation and Selection with Chain-of-Thought from Labeled Data. arXiv preprint arXiv:2302.12822

    Pith/arXiv arXiv 2023

  30. [30]

    Song, R.; Giunchiglia, F.; Li, Y.; Tian, M.; and Xu, H. 2024. Tacit: A target-agnostic feature disentanglement framework for cross-domain text classification. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 38, 18999--19007

    2024

  31. [31]

    Sutton, R. S. 1988. Learning to predict by the methods of temporal differences. Machine learning, 3: 9--44

    1988

  32. [32]

    Tan, Q.; He, R.; Bing, L.; and Ng, H. T. 2022. Domain generalization for text classification with memory-based supervised contrastive learning. In Proceedings of the 29th International Conference on Computational Linguistics, 6916--6926

    2022

  33. [33]

    Tang, H.; Liu, J.; Zhao, M.; and Gong, X. 2020. Progressive layered extraction (ple): A novel multi-task learning (mtl) model for personalized recommendations. In Proceedings of the 14th ACM conference on recommender systems, 269--278

    2020

  34. [34]

    Verma, V.; Fleisig, E.; Tomlin, N.; and Klein, D. 2024. 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), 1702--1717

    2024

  35. [35]

    Wang, P.; Li, L.; Ren, K.; Jiang, B.; Zhang, D.; and Qiu, X. 2023. Seq XGPT : Sentence-Level AI -Generated Text Detection. In The 2023 Conference on Empirical Methods in Natural Language Processing

    2023

  36. [36]

    Wang, Y.; Feng, S.; Hou, A.; Pu, X.; Shen, C.; Liu, X.; Tsvetkov, Y.; and He, T. 2024. Stumbling Blocks: Stress Testing the Robustness of Machine-Generated Text Detectors Under Attacks. In Ku, L.-W.; Martins, A.; and Srikumar, V., eds., Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), 2894--2...

    2024

  37. [37]

    Zellers, R.; Holtzman, A.; Rashkin, H.; Bisk, Y.; Farhadi, A.; Roesner, F.; and Choi, Y. 2019. Defending against neural fake news. Advances in neural information processing systems, 32

    2019

  38. [38]

    Zhong, T.; Chi, Z.; Gu, L.; Wang, Y.; Yu, Y.; and Tang, J. 2022. Meta-dmoe: Adapting to domain shift by meta-distillation from mixture-of-experts. Advances in Neural Information Processing Systems, 35: 22243--22257

    2022