.png){kind=icon}
.png){kind=icon}
Return
EnCur: Curriculum-based in-context learning with structural encoding for code time complexity prediction
.png){kind=icon}
PRE
AI
DOI：10.1016/j.eswa.2025.129094.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
• Structural encoding with ANTLR abstracts code syntax by masking variable and function names to capture underlying algorithmic patterns. • Curriculum learning orders code examples from simple to complex, enabling step-by-step training for time complexity prediction. • Iterative feedback refines model outputs through continuous error correction, improving prediction accuracy and F1 scores. • Experiments with GPT-3.5, GPT-4o, and GPT-4o-mini show that EnCur outperforms stateof- the-art in-context learning techniques for code time complexity prediction.
Keywords:
structural encoding
curriculum learning
iterative feedback
time complexity prediction
in-context learning
Journal
IF:
7.5
Papers:
2.9W
Citations:
10.2W