Search Field
2024
Oedingen, Marc; Engelhardt, Raphael C.; Denz, Robin; Hammer, Maximilian; Konen, Wolfgang
ChatGPT Code Detection: Techniques for Uncovering the Source of Code Artikel
In: arXiv preprint arXiv:2405.15512, 2024.
@article{Oedingen2024,
title = {ChatGPT Code Detection: Techniques for Uncovering the Source of Code},
author = {Marc Oedingen and Raphael C. Engelhardt and Robin Denz and Maximilian Hammer and Wolfgang Konen},
url = {https://arxiv.org/abs/2405.15512},
year = {2024},
date = {2024-01-01},
journal = {arXiv preprint arXiv:2405.15512},
keywords = {AI, ChatGPT, Code Detection, Large Language Models, machine learning},
pubstate = {published},
tppubtype = {article}
}
Oedingen, Marc; Engelhardt, Raphael C.; Denz, Robin; Hammer, Maximilian; Konen, Wolfgang
ChatGPT Code Detection: Techniques for Uncovering the Source of Code Artikel
In: AI, Bd. 5, Nr. 3, S. 1066–1094, 2024, ISSN: 2673-2688.
@article{Oedingen2024a,
title = {ChatGPT Code Detection: Techniques for Uncovering the Source of Code},
author = {Marc Oedingen and Raphael C. Engelhardt and Robin Denz and Maximilian Hammer and Wolfgang Konen},
url = {https://www.mdpi.com/2673-2688/5/3/53},
doi = {10.3390/ai5030053},
issn = {2673-2688},
year = {2024},
date = {2024-01-01},
urldate = {2024-01-01},
journal = {AI},
volume = {5},
number = {3},
pages = {1066–1094},
abstract = {In recent times, large language models (LLMs) have made significant strides in generating computer code, blurring the lines between code created by humans and code produced by artificial intelligence (AI). As these technologies evolve rapidly, it is crucial to explore how they influence code generation, especially given the risk of misuse in areas such as higher education. The present paper explores this issue by using advanced classification techniques to differentiate between code written by humans and code generated by ChatGPT, a type of LLM. We employ a new approach that combines powerful embedding features (black-box) with supervised learning algorithms including Deep Neural Networks, Random Forests, and Extreme Gradient Boosting to achieve this differentiation with an impressive accuracy of 98%. For the successful combinations, we also examine their model calibration, showing that some of the models are extremely well calibrated. Additionally, we present white-box features and an interpretable Bayes classifier to elucidate critical differences between the code sources, enhancing the explainability and transparency of our approach. Both approaches work well, but provide at most 85–88% accuracy. Tests on a small sample of untrained humans suggest that humans do not solve the task much better than random guessing. This study is crucial in understanding and mitigating the potential risks associated with using AI in code generation, particularly in the context of higher education, software development, and competitive programming.},
keywords = {AI, ChatGPT, Code Detection, Large Language Models, machine learning},
pubstate = {published},
tppubtype = {article}
}
In recent times, large language models (LLMs) have made significant strides in generating computer code, blurring the lines between code created by humans and code produced by artificial intelligence (AI). As these technologies evolve rapidly, it is crucial to explore how they influence code generation, especially given the risk of misuse in areas such as higher education. The present paper explores this issue by using advanced classification techniques to differentiate between code written by humans and code generated by ChatGPT, a type of LLM. We employ a new approach that combines powerful embedding features (black-box) with supervised learning algorithms including Deep Neural Networks, Random Forests, and Extreme Gradient Boosting to achieve this differentiation with an impressive accuracy of 98%. For the successful combinations, we also examine their model calibration, showing that some of the models are extremely well calibrated. Additionally, we present white-box features and an interpretable Bayes classifier to elucidate critical differences between the code sources, enhancing the explainability and transparency of our approach. Both approaches work well, but provide at most 85–88% accuracy. Tests on a small sample of untrained humans suggest that humans do not solve the task much better than random guessing. This study is crucial in understanding and mitigating the potential risks associated with using AI in code generation, particularly in the context of higher education, software development, and competitive programming.
Engelhardt, Raphael C; Meinen, Marcel J; Lange, Moritz; Wiskott, Laurenz; Konen, Wolfgang
Putting the Iterative Training of Decision Trees to the Test on a Real-World Robotic Task Artikel
In: arXiv preprint arXiv:2412.04974, 2024.
@article{Engelhardt2024,
title = {Putting the Iterative Training of Decision Trees to the Test on a Real-World Robotic Task},
author = {Raphael C Engelhardt and Marcel J Meinen and Moritz Lange and Laurenz Wiskott and Wolfgang Konen},
url = {https://arxiv.org/pdf/2412.04974},
year = {2024},
date = {2024-01-01},
urldate = {2024-01-01},
journal = {arXiv preprint arXiv:2412.04974},
keywords = {AI, decision trees, deep learning, explainable AI, Reinforcement learning, RL3},
pubstate = {published},
tppubtype = {article}
}
2023
Engelhardt, Raphael C; Oedingen, Marc; Lange, Moritz; Wiskott, Laurenz; Konen, Wolfgang
Iterative Oblique Decision Trees Deliver Explainable RL Models Artikel
In: Algorithms, Bd. 16, Nr. 6, S. 282, 2023.
@article{Engelhardt2023,
title = {Iterative Oblique Decision Trees Deliver Explainable RL Models},
author = {Raphael C Engelhardt and Marc Oedingen and Moritz Lange and Laurenz Wiskott and Wolfgang Konen},
url = {https://www.mdpi.com/1999-4893/16/6/282},
year = {2023},
date = {2023-01-01},
urldate = {2023-01-01},
journal = {Algorithms},
volume = {16},
number = {6},
pages = {282},
publisher = {MDPI},
keywords = {AI, decision trees, deep learning, explainable AI, Reinforcement learning, RL3},
pubstate = {published},
tppubtype = {article}
}
2022
Weitz, Ann
Untersuchung von selbstlernenden Reinforcement Learning Agenten im computergenerierten Spiel Yavalath Abschlussarbeit
TH Köln – University of Applied Sciences, 2022, (Bachelor thesis).
@mastersthesis{Weitz2022b,
title = {Untersuchung von selbstlernenden Reinforcement Learning Agenten im computergenerierten Spiel Yavalath},
author = {Ann Weitz},
url = {https://www.gm.fh-koeln.de/~konen/research/PaperPDF/BA-Weitz-final-2022.pdf},
year = {2022},
date = {2022-05-01},
school = {TH Köln – University of Applied Sciences},
note = {Bachelor thesis},
keywords = {AI, BT-MT, Game Learning, Reinforcement learning},
pubstate = {published},
tppubtype = {mastersthesis}
}
Weitz, Ann
Entwicklung einer allgemeinen Schnittstelle zwischen Ludii und dem GBG Framework Forschungsbericht
2022, (Praxisprojekt).
@techreport{Weitz2022,
title = {Entwicklung einer allgemeinen Schnittstelle zwischen Ludii und dem GBG Framework},
author = {Ann Weitz},
url = {https://www.gm.fh-koeln.de/~konen/research/PaperPDF/PP-Doku-Weitz-2022-02.pdf},
year = {2022},
date = {2022-02-01},
school = {TH Köln – University of Applied Sciences},
note = {Praxisprojekt},
keywords = {AI, BT-MT, Game Learning, Ludii, Reinforcement learning},
pubstate = {published},
tppubtype = {techreport}
}
Cöln, Julian
KI-Konzepte für das Erlernen nicht-deterministischer Spiele am Beispiel von "EinStein würfelt nicht!" Abschlussarbeit
TH Köln – University of Applied Sciences, 2022, (Bachelor thesis).
@mastersthesis{Coeln2022,
title = {KI-Konzepte für das Erlernen nicht-deterministischer Spiele am Beispiel von "EinStein würfelt nicht!"},
author = {Julian Cöln},
url = {https://www.gm.fh-koeln.de/~konen/research/PaperPDF/BA_Julian_Coeln2021-final.pdf},
year = {2022},
date = {2022-02-01},
school = {TH Köln – University of Applied Sciences},
note = {Bachelor thesis},
keywords = {AI, BT-MT, Game Learning, GBG, Reinforcement learning},
pubstate = {published},
tppubtype = {mastersthesis}
}
Scheiermann, Johannes; Konen, Wolfgang
AlphaZero-Inspired Game Learning: Faster Training by Using MCTS Only at Test Time Artikel
In: IEEE Transactions on Games, 2022.
@article{Scheier2022,
title = {AlphaZero-Inspired Game Learning: Faster Training by Using MCTS Only at Test Time},
author = {Johannes Scheiermann and Wolfgang Konen},
url = {https://ieeexplore.ieee.org/document/9893320},
doi = {10.1109/TG.2022.3206733},
year = {2022},
date = {2022-01-01},
urldate = {2022-01-01},
journal = {IEEE Transactions on Games},
keywords = {AI, deep learning, Game Learning, GBG, learning, optimization, Reinforcement learning},
pubstate = {published},
tppubtype = {article}
}
2021
Meissner, Simon
Untersuchung des Spiel- und Lernerfolgs künstlicher Intelligenzen für ein nichtdeterministisches Spiel mit imperfekten Informationen: Blackjack in der Game-Learning-Umgebung ’General Board Game’ (GBG) Abschlussarbeit
TH Köln – University of Applied Sciences, 2021, (Bachelor thesis).
@mastersthesis{Meissner2021,
title = {Untersuchung des Spiel- und Lernerfolgs künstlicher Intelligenzen für ein nichtdeterministisches Spiel mit imperfekten Informationen: Blackjack in der Game-Learning-Umgebung ’General Board Game’ (GBG)},
author = {Simon Meissner},
url = {https://www.gm.fh-koeln.de/~konen/research/PaperPDF/BA-Meissner-final-2021.pdf},
year = {2021},
date = {2021-12-01},
school = {TH Köln – University of Applied Sciences},
note = {Bachelor thesis},
keywords = {AI, BT-MT, Game Learning, GBG, machine learning},
pubstate = {published},
tppubtype = {mastersthesis}
}
Zeh, Tim
Untersuchung von allgemeinen KI-Agenten für das Spiel Poker im General Board Games Framework Abschlussarbeit
TH Köln – University of Applied Sciences, 2021, (Master thesis).
@mastersthesis{Zeh2021,
title = {Untersuchung von allgemeinen KI-Agenten für das Spiel Poker im General Board Games Framework},
author = {Tim Zeh},
url = {https://www.gm.fh-koeln.de/~konen/research/PaperPDF/MA_Zeh_final_Poker-GBG-2021.pdf},
year = {2021},
date = {2021-07-01},
school = {TH Köln – University of Applied Sciences},
note = {Master thesis},
keywords = {AI, BT-MT, Game Learning, GBG, machine learning},
pubstate = {published},
tppubtype = {mastersthesis}
}
2020
Scheiermann, Johannes
Sind (trainierte) General-Purpose-RL-Agenten im Brettspiel Othello stärker als (untrainierte) General-Game-Playing Agenten? Forschungsbericht
TH Köln, Institut für Informatik 2020, (Praxisprojekt).
@techreport{Scheier2020,
title = {Sind (trainierte) General-Purpose-RL-Agenten im Brettspiel Othello stärker als (untrainierte) General-Game-Playing Agenten?},
author = {Johannes Scheiermann},
url = {http://www.gm.fh-koeln.de/~konen/research/PaperPDF/INF-Prj-Scheiermann-2020-08.pdf},
year = {2020},
date = {2020-01-01},
institution = {TH Köln, Institut für Informatik},
note = {Praxisprojekt},
keywords = {AI, BT-MT, Game Learning, GBG, machine learning, Reinforcement learning},
pubstate = {published},
tppubtype = {techreport}
}
Scheiermann, Johannes
AlphaZero-inspirierte KI-Agenten im General Board Game Playing Abschlussarbeit
TH Köln -- University of Applied Sciences, 2020, (Bachelor thesis).
@mastersthesis{Scheier2020b,
title = {AlphaZero-inspirierte KI-Agenten im General Board Game Playing},
author = {Johannes Scheiermann},
url = {http://www.gm.fh-koeln.de/~konen/research/PaperPDF/BA_Scheiermann_final.pdf},
year = {2020},
date = {2020-01-01},
school = {TH Köln -- University of Applied Sciences},
note = {Bachelor thesis},
keywords = {AI, BT-MT, Game Learning, GBG, machine learning, Reinforcement learning},
pubstate = {published},
tppubtype = {mastersthesis}
}
2019
Cöln, Julian; Dittmar, Yannick
Untersuchung von KI Agenten im Spiel Othello Forschungsbericht
TH Köln, Institut für Informatik 2019.
@techreport{Cöln2019,
title = {Untersuchung von KI Agenten im Spiel Othello},
author = {Julian Cöln and Yannick Dittmar},
url = {http://www.gm.fh-koeln.de/~konen/research/PaperPDF/INF-Proj-DittmarCoeln-2019-12.pdf},
year = {2019},
date = {2019-12-01},
institution = {TH Köln, Institut für Informatik},
keywords = {AI, BT-MT, Game Learning, GBG, machine learning, Reinforcement learning},
pubstate = {published},
tppubtype = {techreport}
}
Search Field
13 Einträge « ‹ 1 von 2
› » 1.
Oedingen, Marc; Engelhardt, Raphael C.; Denz, Robin; Hammer, Maximilian; Konen, Wolfgang
ChatGPT Code Detection: Techniques for Uncovering the Source of Code Artikel
In: arXiv preprint arXiv:2405.15512, 2024.
@article{Oedingen2024,
title = {ChatGPT Code Detection: Techniques for Uncovering the Source of Code},
author = {Marc Oedingen and Raphael C. Engelhardt and Robin Denz and Maximilian Hammer and Wolfgang Konen},
url = {https://arxiv.org/abs/2405.15512},
year = {2024},
date = {2024-01-01},
journal = {arXiv preprint arXiv:2405.15512},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
2.
Oedingen, Marc; Engelhardt, Raphael C.; Denz, Robin; Hammer, Maximilian; Konen, Wolfgang
ChatGPT Code Detection: Techniques for Uncovering the Source of Code Artikel
In: AI, Bd. 5, Nr. 3, S. 1066–1094, 2024, ISSN: 2673-2688.
@article{Oedingen2024a,
title = {ChatGPT Code Detection: Techniques for Uncovering the Source of Code},
author = {Marc Oedingen and Raphael C. Engelhardt and Robin Denz and Maximilian Hammer and Wolfgang Konen},
url = {https://www.mdpi.com/2673-2688/5/3/53},
doi = {10.3390/ai5030053},
issn = {2673-2688},
year = {2024},
date = {2024-01-01},
urldate = {2024-01-01},
journal = {AI},
volume = {5},
number = {3},
pages = {1066–1094},
abstract = {In recent times, large language models (LLMs) have made significant strides in generating computer code, blurring the lines between code created by humans and code produced by artificial intelligence (AI). As these technologies evolve rapidly, it is crucial to explore how they influence code generation, especially given the risk of misuse in areas such as higher education. The present paper explores this issue by using advanced classification techniques to differentiate between code written by humans and code generated by ChatGPT, a type of LLM. We employ a new approach that combines powerful embedding features (black-box) with supervised learning algorithms including Deep Neural Networks, Random Forests, and Extreme Gradient Boosting to achieve this differentiation with an impressive accuracy of 98%. For the successful combinations, we also examine their model calibration, showing that some of the models are extremely well calibrated. Additionally, we present white-box features and an interpretable Bayes classifier to elucidate critical differences between the code sources, enhancing the explainability and transparency of our approach. Both approaches work well, but provide at most 85–88% accuracy. Tests on a small sample of untrained humans suggest that humans do not solve the task much better than random guessing. This study is crucial in understanding and mitigating the potential risks associated with using AI in code generation, particularly in the context of higher education, software development, and competitive programming.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
In recent times, large language models (LLMs) have made significant strides in generating computer code, blurring the lines between code created by humans and code produced by artificial intelligence (AI). As these technologies evolve rapidly, it is crucial to explore how they influence code generation, especially given the risk of misuse in areas such as higher education. The present paper explores this issue by using advanced classification techniques to differentiate between code written by humans and code generated by ChatGPT, a type of LLM. We employ a new approach that combines powerful embedding features (black-box) with supervised learning algorithms including Deep Neural Networks, Random Forests, and Extreme Gradient Boosting to achieve this differentiation with an impressive accuracy of 98%. For the successful combinations, we also examine their model calibration, showing that some of the models are extremely well calibrated. Additionally, we present white-box features and an interpretable Bayes classifier to elucidate critical differences between the code sources, enhancing the explainability and transparency of our approach. Both approaches work well, but provide at most 85–88% accuracy. Tests on a small sample of untrained humans suggest that humans do not solve the task much better than random guessing. This study is crucial in understanding and mitigating the potential risks associated with using AI in code generation, particularly in the context of higher education, software development, and competitive programming.
3.
Engelhardt, Raphael C; Meinen, Marcel J; Lange, Moritz; Wiskott, Laurenz; Konen, Wolfgang
Putting the Iterative Training of Decision Trees to the Test on a Real-World Robotic Task Artikel
In: arXiv preprint arXiv:2412.04974, 2024.
@article{Engelhardt2024,
title = {Putting the Iterative Training of Decision Trees to the Test on a Real-World Robotic Task},
author = {Raphael C Engelhardt and Marcel J Meinen and Moritz Lange and Laurenz Wiskott and Wolfgang Konen},
url = {https://arxiv.org/pdf/2412.04974},
year = {2024},
date = {2024-01-01},
urldate = {2024-01-01},
journal = {arXiv preprint arXiv:2412.04974},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
4.
Engelhardt, Raphael C; Oedingen, Marc; Lange, Moritz; Wiskott, Laurenz; Konen, Wolfgang
Iterative Oblique Decision Trees Deliver Explainable RL Models Artikel
In: Algorithms, Bd. 16, Nr. 6, S. 282, 2023.
@article{Engelhardt2023,
title = {Iterative Oblique Decision Trees Deliver Explainable RL Models},
author = {Raphael C Engelhardt and Marc Oedingen and Moritz Lange and Laurenz Wiskott and Wolfgang Konen},
url = {https://www.mdpi.com/1999-4893/16/6/282},
year = {2023},
date = {2023-01-01},
urldate = {2023-01-01},
journal = {Algorithms},
volume = {16},
number = {6},
pages = {282},
publisher = {MDPI},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
5.
Weitz, Ann
Untersuchung von selbstlernenden Reinforcement Learning Agenten im computergenerierten Spiel Yavalath Abschlussarbeit
TH Köln – University of Applied Sciences, 2022, (Bachelor thesis).
@mastersthesis{Weitz2022b,
title = {Untersuchung von selbstlernenden Reinforcement Learning Agenten im computergenerierten Spiel Yavalath},
author = {Ann Weitz},
url = {https://www.gm.fh-koeln.de/~konen/research/PaperPDF/BA-Weitz-final-2022.pdf},
year = {2022},
date = {2022-05-01},
school = {TH Köln – University of Applied Sciences},
note = {Bachelor thesis},
keywords = {},
pubstate = {published},
tppubtype = {mastersthesis}
}
6.
Weitz, Ann
Entwicklung einer allgemeinen Schnittstelle zwischen Ludii und dem GBG Framework Forschungsbericht
2022, (Praxisprojekt).
@techreport{Weitz2022,
title = {Entwicklung einer allgemeinen Schnittstelle zwischen Ludii und dem GBG Framework},
author = {Ann Weitz},
url = {https://www.gm.fh-koeln.de/~konen/research/PaperPDF/PP-Doku-Weitz-2022-02.pdf},
year = {2022},
date = {2022-02-01},
school = {TH Köln – University of Applied Sciences},
note = {Praxisprojekt},
keywords = {},
pubstate = {published},
tppubtype = {techreport}
}
7.
Cöln, Julian
KI-Konzepte für das Erlernen nicht-deterministischer Spiele am Beispiel von "EinStein würfelt nicht!" Abschlussarbeit
TH Köln – University of Applied Sciences, 2022, (Bachelor thesis).
@mastersthesis{Coeln2022,
title = {KI-Konzepte für das Erlernen nicht-deterministischer Spiele am Beispiel von "EinStein würfelt nicht!"},
author = {Julian Cöln},
url = {https://www.gm.fh-koeln.de/~konen/research/PaperPDF/BA_Julian_Coeln2021-final.pdf},
year = {2022},
date = {2022-02-01},
school = {TH Köln – University of Applied Sciences},
note = {Bachelor thesis},
keywords = {},
pubstate = {published},
tppubtype = {mastersthesis}
}
8.
Scheiermann, Johannes; Konen, Wolfgang
AlphaZero-Inspired Game Learning: Faster Training by Using MCTS Only at Test Time Artikel
In: IEEE Transactions on Games, 2022.
@article{Scheier2022,
title = {AlphaZero-Inspired Game Learning: Faster Training by Using MCTS Only at Test Time},
author = {Johannes Scheiermann and Wolfgang Konen},
url = {https://ieeexplore.ieee.org/document/9893320},
doi = {10.1109/TG.2022.3206733},
year = {2022},
date = {2022-01-01},
urldate = {2022-01-01},
journal = {IEEE Transactions on Games},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
9.
Meissner, Simon
Untersuchung des Spiel- und Lernerfolgs künstlicher Intelligenzen für ein nichtdeterministisches Spiel mit imperfekten Informationen: Blackjack in der Game-Learning-Umgebung ’General Board Game’ (GBG) Abschlussarbeit
TH Köln – University of Applied Sciences, 2021, (Bachelor thesis).
@mastersthesis{Meissner2021,
title = {Untersuchung des Spiel- und Lernerfolgs künstlicher Intelligenzen für ein nichtdeterministisches Spiel mit imperfekten Informationen: Blackjack in der Game-Learning-Umgebung ’General Board Game’ (GBG)},
author = {Simon Meissner},
url = {https://www.gm.fh-koeln.de/~konen/research/PaperPDF/BA-Meissner-final-2021.pdf},
year = {2021},
date = {2021-12-01},
school = {TH Köln – University of Applied Sciences},
note = {Bachelor thesis},
keywords = {},
pubstate = {published},
tppubtype = {mastersthesis}
}
10.
Zeh, Tim
Untersuchung von allgemeinen KI-Agenten für das Spiel Poker im General Board Games Framework Abschlussarbeit
TH Köln – University of Applied Sciences, 2021, (Master thesis).
@mastersthesis{Zeh2021,
title = {Untersuchung von allgemeinen KI-Agenten für das Spiel Poker im General Board Games Framework},
author = {Tim Zeh},
url = {https://www.gm.fh-koeln.de/~konen/research/PaperPDF/MA_Zeh_final_Poker-GBG-2021.pdf},
year = {2021},
date = {2021-07-01},
school = {TH Köln – University of Applied Sciences},
note = {Master thesis},
keywords = {},
pubstate = {published},
tppubtype = {mastersthesis}
}
13 Einträge « ‹ 1 von 2
› »
