CodeMixBench adds code-mixed prompt variants to BigCodeBench and shows that code-mixed inputs consistently reduce Pass@1 performance of code LLMs.
Mipe: A metric independent pipeline for effective code-mixed nlg evaluation
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
CodeMixBench: Evaluating Large Language Models on Code Generation with Code-Mixed Prompts
CodeMixBench adds code-mixed prompt variants to BigCodeBench and shows that code-mixed inputs consistently reduce Pass@1 performance of code LLMs.