This research introduces a novel repository-level benchmark for evaluating code translation capabilities of large language models, specifically targeting Rust as a destination language.

• First benchmark to evaluate code translation with full repository context and dependencies
• Addresses significant gap in existing benchmarks that only test isolated function-level translations
• Shows current limitations of LLMs when translating to newer languages like Rust in complex scenarios
• Provides insights for improving LLMs for real-world software engineering tasks

This work matters for engineering teams seeking to modernize codebases or migrate to Rust, highlighting both the potential and current limitations of automated translation tools in production environments.

Repository-level Code Translation Benchmark Targeting Rust