A hybrid CodeBERT + GPT-3.5 code completion model reportedly outperforms its components on CodeXGLUE-derived Python data, with no reproducible artifacts.
The model demonstrates remarkable accuracy improvements, achieving an F1-Score of 0.91, which represents a substantial 13.75% enhancement compared to the baseline model
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.DC 1years
2025 1verdicts
REJECT 1representative citing papers
citing papers explorer
-
Design and Implementation of Code Completion System Based on LLM and CodeBERT Hybrid Subsystem
A hybrid CodeBERT + GPT-3.5 code completion model reportedly outperforms its components on CodeXGLUE-derived Python data, with no reproducible artifacts.