Code Clarity

CodeClarity is a research initiative focused on evaluating large language models for multilingual code understanding and documentation.

The project introduces CodeClarity-Bench, a benchmark for evaluating code summarization across multiple programming languages and natural languages.

The framework evaluates both traditional metrics (BLEU, ROUGE-L, METEOR, ChrF++, BERTScore, COMET) and LLM-as-a-judge evaluation methods.

Resources: • Dataset: https://huggingface.co/CodeClarity/CodeClarity-Bench
• Code: https://github.com/MadhuNimmo/CodeClarity

The benchmark is introduced in the LREC-COLING 2026 paper: "CodeClarity: A Framework and Benchmark for Evaluating Multilingual Code Summarization."