Francisco Ortin
@franciscoortin.bsky.social
Full professor @ University of Oviedo
https://www.reflection.uniovi.es/ortin
https://www.reflection.uniovi.es/ortin
We propose a CNN + RNN model to detect cheating in online programming exams via screenshot sequences. It achieves 95.18% accuracy & 94.2% F2-score. Data augmentation + class weights improved results; transfer learning & alt. loss functions gave no gains.
September 9, 2025 at 11:00 AM
We propose a CNN + RNN model to detect cheating in online programming exams via screenshot sequences. It achieves 95.18% accuracy & 94.2% F2-score. Data augmentation + class weights improved results; transfer learning & alt. loss functions gave no gains.
PlangRec is a character-level deep-learning model that predicts the language of a line of code. It provides 95% accuracy and macro F1-score to classify 21 different languages, outperforming state-of-the-art systems for classifying code snippets.
doi.org/10.1016/j.fu...
doi.org/10.1016/j.fu...
Redirecting
doi.org
March 14, 2025 at 8:55 AM
PlangRec is a character-level deep-learning model that predicts the language of a line of code. It provides 95% accuracy and macro F1-score to classify 21 different languages, outperforming state-of-the-art systems for classifying code snippets.
doi.org/10.1016/j.fu...
doi.org/10.1016/j.fu...
CLAVE is a one-shot deep-learning model for source code authorship verification. Contrastive learning is used to learn stylometric representations. CLAVE outperforms the state-of-the-art systems with 0.9782 AUC.
doi.org/10.1016/j.ip...
doi.org/10.1016/j.ip...
Francisco Ortin on X: "CLAVE is a one-shot deep-learning model for source code authorship verification. Contrastive learning is used to learn stylometric representations. CLAVE outperforms the state-of-the-art systems with 0.9782 AUC. https://t.co/qjttIAFSQe" / X
CLAVE is a one-shot deep-learning model for source code authorship verification. Contrastive learning is used to learn stylometric representations. CLAVE outperforms the state-of-the-art systems with 0.9782 AUC. https://t.co/qjttIAFSQe
x.com
March 14, 2025 at 8:48 AM
CLAVE is a one-shot deep-learning model for source code authorship verification. Contrastive learning is used to learn stylometric representations. CLAVE outperforms the state-of-the-art systems with 0.9782 AUC.
doi.org/10.1016/j.ip...
doi.org/10.1016/j.ip...