Feeding LLMs their own code's complexity metrics as feedback modestly improves Pass@1 on some benchmarks, especially for weaker models like GPT-3.5 Turbo.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.SE 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Enhancing LLM-Based Code Generation with Complexity Metrics: A Feedback-Driven Approach
Feeding LLMs their own code's complexity metrics as feedback modestly improves Pass@1 on some benchmarks, especially for weaker models like GPT-3.5 Turbo.