Pith. sign in

REVIEW 5 cited by

MAGE: Machine-generated Text Detection in the Wild

Not yet reviewed by Pith; the record is open.

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

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2305.13242 v3 pith:6EHEJH5L submitted 2023-05-22 cs.CL

MAGE: Machine-generated Text Detection in the Wild

Yafu Li , Qintong Li , Leyang Cui , Wei Bi , Zhilin Wang , Longyue Wang , Linyi Yang , Shuming Shi
show 1 more author
Yue Zhang
This is my paper
classification cs.CL
keywords textschallengesdetectionllmsscenariostextdetectordomains
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Large language models (LLMs) have achieved human-level text generation, emphasizing the need for effective AI-generated text detection to mitigate risks like the spread of fake news and plagiarism. Existing research has been constrained by evaluating detection methods on specific domains or particular language models. In practical scenarios, however, the detector faces texts from various domains or LLMs without knowing their sources. To this end, we build a comprehensive testbed by gathering texts from diverse human writings and texts generated by different LLMs. Empirical results show challenges in distinguishing machine-generated texts from human-authored ones across various scenarios, especially out-of-distribution. These challenges are due to the decreasing linguistic distinctions between the two sources. Despite challenges, the top-performing detector can identify 86.54% out-of-domain texts generated by a new LLM, indicating the feasibility for application scenarios. We release our resources at https://github.com/yafuly/MAGE.

discussion (0)

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

Forward citations

Cited by 5 Pith papers

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

  1. MAGA-Bench: Machine-Augment-Generated Text via Alignment Detection Benchmark

    cs.CL 2026-01 conditional novelty 6.0

    Adding human-alignment augmentation (roleplaying, BPO, self-refine, RLDF) to machine-generated text both fools existing detectors and improves the generalization of detectors fine-tuned on it.

  2. Monitoring AI-Modified Content at Scale: A Case Study on the Impact of ChatGPT on AI Conference Peer Reviews

    cs.CL 2024-03 unverdicted novelty 6.0

    A maximum likelihood model estimates 6.5-16.9% of peer-review text at ICLR 2024, NeurIPS 2023, CoRL 2023 and EMNLP 2023 was substantially modified by LLMs, with elevated rates in low-confidence and deadline-close submissions.

  3. Can AI-Generated Text be Reliably Detected?

    cs.CL 2023-03 unverdicted novelty 6.0

    Recursive paraphrasing attacks substantially lower detection rates for multiple AI text detectors with only minor quality loss, while a theoretical analysis ties best-case AUROC to total variation distance between hum...

  4. C-ReD: A Comprehensive Chinese Benchmark for AI-Generated Text Detection Derived from Real-World Prompts

    cs.CL 2026-04 unverdicted novelty 5.0

    C-ReD is a new Chinese benchmark for AI-generated text detection built from diverse real-world prompts to improve in-domain performance and generalization to unseen models and datasets.

  5. C-ReD: A Comprehensive Chinese Benchmark for AI-Generated Text Detection Derived from Real-World Prompts

    cs.CL 2026-04 unverdicted novelty 4.0

    C-ReD is a Chinese AI-text detection benchmark built from diverse real-world prompts and multiple LLMs that shows strong in-domain performance and generalization to unseen models and external datasets.