Papers
Topics
Authors
Recent
Gemini 2.5 Flash
Gemini 2.5 Flash
119 tokens/sec
GPT-4o
56 tokens/sec
Gemini 2.5 Pro Pro
43 tokens/sec
o3 Pro
6 tokens/sec
GPT-4.1 Pro
47 tokens/sec
DeepSeek R1 via Azure Pro
28 tokens/sec
2000 character limit reached

LLM-as-a-Coauthor: Can Mixed Human-Written and Machine-Generated Text Be Detected? (2401.05952v2)

Published 11 Jan 2024 in cs.CL

Abstract: With the rapid development and widespread application of LLMs, the use of Machine-Generated Text (MGT) has become increasingly common, bringing with it potential risks, especially in terms of quality and integrity in fields like news, education, and science. Current research mainly focuses on purely MGT detection without adequately addressing mixed scenarios, including AI-revised Human-Written Text (HWT) or human-revised MGT. To tackle this challenge, we define mixtext, a form of mixed text involving both AI and human-generated content. Then, we introduce MixSet, the first dataset dedicated to studying these mixtext scenarios. Leveraging MixSet, we executed comprehensive experiments to assess the efficacy of prevalent MGT detectors in handling mixtext situations, evaluating their performance in terms of effectiveness, robustness, and generalization. Our findings reveal that existing detectors struggle to identify mixtext, particularly in dealing with subtle modifications and style adaptability. This research underscores the urgent need for more fine-grain detectors tailored for mixtext, offering valuable insights for future research. Code and Models are available at https://github.com/Dongping-Chen/MixSet.

Definition Search Book Streamline Icon: https://streamlinehq.com
References (83)
  1. Google translate. https://translate.google.com/.
  2. Grammarly. https://www.grammarly.com/.
  3. Youdao translate. http://fanyi.youdao.com/.
  4. AIContentfy. 2023. Chatgpt in the gaming industry: Enhancing storytelling and interaction. https://aicontentfy.com/en/blog/chatgpt-in-gaming-industry-enhancing-storytelling-and-interaction.
  5. Gpt4all: Training an assistant-style chatbot with large scale data distillation from gpt-3.5-turbo. GitHub.
  6. Real or fake? learning to discriminate machine from human generated text. arXiv preprint arXiv:1906.03351.
  7. Fast-detectgpt: Efficient zero-shot detection of machine-generated text via conditional probability curvature. arXiv preprint arXiv:2310.05130.
  8. Gandhi Gram Bhudghar. 2023. Ai text converter. https://aitextconverter.com/.
  9. Natural language processing with python: Analyzing text with the natural language toolkit. http://nltk.org/.
  10. Gpt-neox-20b: An open-source autoregressive language model. arXiv preprint arXiv:2204.06745.
  11. Xuhang Chen. 2023. Gpt academic prompt. https://github.com/xuhangc/ChatGPT-Academic-Prompt.
  12. Gpt-sentinel: Distinguishing human and chatgpt generated content. arXiv preprint arXiv:2305.07969.
  13. Vicuna: An open-source chatbot impressing gpt-4 with 90%* chatgpt quality.
  14. Jon Christian. 2023. Cnet secretly used ai on articles that didn’t disclose that fact, staff say. https://futurism.com/cnet-ai-articles-label.
  15. CMU. 2015. Enron email dataset. https://www.cs.cmu.edu/~enron/.
  16. Free dolly: Introducing the world’s first truly open instruction-tuned llm. https://www.databricks.com/blog/2023/04/12/dolly-first-open-commercially-viable-instruction-tuned-llm.
  17. Glm: General language model pretraining with autoregressive blank infilling. In Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 320–335.
  18. Holly Else. 2023. Abstracts written by chatgpt fool scientists. Nature, 613(7944):423–423.
  19. Protoformer: Embedding prototypes for transformers. In Advances in Knowledge Discovery and Data Mining: 26th Pacific-Asia Conference, PAKDD 2022, Chengdu, China, May 16–19, 2022, Proceedings, Part I, pages 447–458.
  20. Gltr: Statistical detection and visualization of generated text. arXiv preprint arXiv:1906.04043.
  21. Katy Ilonka Gero and Lydia B Chilton. 2019. Metaphoria: An algorithmic companion for metaphor creation. In Proceedings of the 2019 CHI conference on human factors in computing systems, pages 1–12.
  22. Sparks: Inspiration for science writing using language models. pages 1002–1019.
  23. Towards possibilities & impossibilities of ai-generated text detection: A survey. arXiv preprint arXiv:2310.15264.
  24. A survey of adversarial defenses and robustness in nlp. ACM Computing Surveys, 55(14s):1–39.
  25. Derek Greene and Pádraig Cunningham. 2006. Practical solutions to the problem of diagonal dominance in kernel document clustering. In Proceedings of the 23rd international conference on Machine learning, pages 377–384.
  26. On the learnability of watermarks for language models.
  27. Marci Guerra. 2023. Chat gpt for journalism: Revolutionizing the future of reporting. https://brandalytics.co/chat-gpt-for-journalism/.
  28. How close is chatgpt to human experts? comparison corpus, evaluation, and detection. arXiv preprint arXiv:2301.07597.
  29. Zhen Guo and Shangdi Yu. 2023. Authentigpt: Detecting machine-generated text via black-box language models denoising. arXiv preprint arXiv:2311.07700.
  30. news-please: A generic news crawler and extractor.
  31. Mgtbench: Benchmarking machine-generated text detection. arXiv preprint arXiv:2303.14822.
  32. Will Douglas Heavenarchive. 2023. Chatgpt is going to change education, not destroy it. https://www.technologyreview.com/2023/04/06/1071059/chatgpt-change-not-destroy-education-openai/.
  33. Radar: Robust ai-text detection via adversarial learning. arXiv preprint arXiv:2307.03838.
  34. Hugging Face. 2023. Open llm leaderboard. https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard.
  35. Is chatgpt a “fire of prometheus” for non-native english-speaking researchers in academic writing? Korean Journal of Radiology, 24(10):952.
  36. Automatic detection of generated text is easiest when humans are fooled. arXiv preprint arXiv:1911.00650.
  37. Pubmedqa: A dataset for biomedical research question answering. arXiv preprint arXiv:1909.06146.
  38. A watermark for large language models. arXiv preprint arXiv:2301.10226.
  39. The narrativeqa reading comprehension challenge. Transactions of the Association for Computational Linguistics, 6:317–328.
  40. How you prompt matters! even task-oriented constraints in instructions affect llm-generated text detection. arXiv preprint arXiv:2311.08369.
  41. Outfox: Llm-generated essay detection through in-context learning with adversarially generated examples. arXiv preprint arXiv:2307.11729.
  42. Paraphrasing evades detectors of ai-generated text, but retrieval is an effective defense. arXiv preprint arXiv:2303.13408.
  43. Vladimir Iosifovich Levenshtein. 1966. Binary codes capable of correcting deletions, insertions and reversals. Soviet Physics Doklady, 10(8):707–710.
  44. Truthfulqa: Measuring how models mimic human falsehoods. arXiv preprint arXiv:2109.07958.
  45. TruthfulQA: Measuring how models mimic human falsehoods. pages 3214–3252, Dublin, Ireland. Association for Computational Linguistics.
  46. Ilya Loshchilov and Frank Hutter. 2019. Decoupled weight decay regularization.
  47. Smaller language models are better black-box machine-generated text detectors. arXiv preprint arXiv:2305.09859.
  48. Detectgpt: Zero-shot machine-generated text detection using probability curvature. arXiv preprint arXiv:2301.11305.
  49. Covid-qa: A question answering dataset for covid-19. In Proceedings of the 1st Workshop on NLP for COVID-19 at ACL 2020.
  50. Crosslingual generalization through multitask finetuning. arXiv preprint arXiv:2211.01786.
  51. Najzeko. 2021. Steam reviews dataset 2021.
  52. Multi-style generative reading comprehension. pages 2273–2284, Florence, Italy. Association for Computational Linguistics.
  53. OpenAI. 2022. Openai models - gpt3.5. https://platform.openai.com/docs/models/gpt-3-5.
  54. OpenAI. 2023a. Ai text classifier. https://beta.openai.com/ai-text-classifier.
  55. OpenAI. 2023b. Gpt-4 technical report.
  56. Language models are unsupervised multitask learners.
  57. Exploring the limits of transfer learning with a unified text-to-text transformer. The Journal of Machine Learning Research, 21(1):5485–5551.
  58. SQuAD: 100,000+ questions for machine comprehension of text. pages 2383–2392, Austin, Texas. Association for Computational Linguistics.
  59. Squad: 100,000+ questions for machine comprehension of text. arXiv preprint arXiv:1606.05250.
  60. Data augmentation can improve robustness. Advances in Neural Information Processing Systems, 34:29935–29948.
  61. Can ai-generated text be reliably detected? arXiv preprint arXiv:2303.11156.
  62. Effects of age and gender on blogging. In AAAI spring symposium: Computational approaches to analyzing weblogs, volume 6, pages 199–205.
  63. Claude Elwood Shannon. 1948. A mathematical theory of communication. The Bell system technical journal, 27(3):379–423.
  64. Rewritelm: An instruction-tuned large language model for text rewriting. arXiv preprint arXiv:2305.15685.
  65. Chatgpt: More than a weapon of mass deception, ethical challenges and responses from the human-centered artificial intelligence (hcai) perspective. arXiv preprint arXiv:2304.11215.
  66. Release strategies and the social impacts of language models. arXiv preprint arXiv:1908.09203.
  67. StabilityAI. 2023. Stablelm.
  68. Detectllm: Leveraging log rank information for zero-shot detection of machine-generated text. arXiv preprint arXiv:2306.05540.
  69. TheDataBeast. 2021. Ted talk transcripts (2006 - 2021).
  70. Edward Tian. 2023. Gptzero: An ai text detector. https://gptzero.me/.
  71. Llama: Open and efficient foundation language models. arXiv preprint arXiv:2302.13971.
  72. Authorship attribution for neural text generation. In Proceedings of the 2020 conference on empirical methods in natural language processing (EMNLP), pages 8384–8395.
  73. Ghostbuster: Detecting text ghostwritten by large language models. arXiv preprint arXiv:2305.15047.
  74. Adversarial glue: A multi-task benchmark for robustness evaluation of language models. arXiv preprint arXiv:2111.02840.
  75. Llmdet: A third party large language models generated text detection tool. In The 2023 Conference on Empirical Methods in Natural Language Processing.
  76. On the generalization of training-based chatgpt detection methods. arXiv preprint arXiv:2310.01307.
  77. Dna-gpt: Divergent n-gram analysis for training-free detection of gpt-generated text. arXiv preprint arXiv:2305.17359.
  78. Zero-shot detection of machine-generated codes. arXiv preprint arXiv:2310.05103.
  79. Xue Ying. 2019. An overview of overfitting and its solutions. In Journal of physics: Conference series, volume 1168, page 022022. IOP Publishing.
  80. Gpt paternity test: Gpt generated text detection with gpt genetic inheritance. arXiv preprint arXiv:2305.12519.
  81. Vanessa Yurkevich. 2023. Experts warn about possible misuse of new ai tool chatgpt. https://www.atlantanewsfirst.com/2023/01/24/experts-warn-about-possible-misuse-new-ai-tool-chatgpt/.
  82. Texygen: A benchmarking platform for text generation models. In The 41st International ACM SIGIR Conference on Research & Development in Information Retrieval, page 1097–1100.
  83. Jey Han Lau Zhuohan Xie, Trevor Cohn. The next chapter: A study of large language models in storytelling. https://aclanthology.org/2023.inlg-main.23/.
User Edit Pencil Streamline Icon: https://streamlinehq.com
Authors (11)
  1. Chujie Gao (9 papers)
  2. Dongping Chen (28 papers)
  3. Qihui Zhang (13 papers)
  4. Yue Huang (171 papers)
  5. Yao Wan (70 papers)
  6. Lichao Sun (186 papers)
  7. Yixin Huang (7 papers)
  8. Zhenyang Sun (1 paper)
  9. Shilin Zhang (8 papers)
  10. Weiye Li (7 papers)
  11. Zhengyan Fu (2 papers)
Citations (15)
View on Semantic Scholar

Summary

We haven't generated a summary for this paper yet.

Ai Sparkles Streamline Icon: https://streamlinehq.com Summarize Now
Github Logo Streamline Icon: https://streamlinehq.com

GitHub

  1. GitHub - Dongping-Chen/MixSet: Official code repository for Mixset. (25 stars)
X Twitter Logo Streamline Icon: https://streamlinehq.com

Tweets