An automated synthetic-data and adaptive-finetuning framework improved 7 of 8 execution-based pass@k comparisons for Java-to-Go and C++-to-Rust, but one pass@1 regressed and no external benchmark was used.
Effective test generation using pre-trained large language models and mutation testing
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.AI 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
ACT: Bridging the Gap in Code Translation through Synthetic Data Generation & Adaptive Training
An automated synthetic-data and adaptive-finetuning framework improved 7 of 8 execution-based pass@k comparisons for Java-to-Go and C++-to-Rust, but one pass@1 regressed and no external benchmark was used.