tongye98/Awesome-Code-Benchmark

Curates and tracks emerging code-specific benchmarks across diverse evaluation dimensions—from code generation and security analysis to repository-level reasoning and multi-turn interactions. Aggregates peer-reviewed research benchmarks with structured categorization by task type, capability tested, and source institution, enabling systematic comparison of LLM performance across hundreds of specialized evaluation datasets. Actively maintains featured benchmark listings covering recent advances in performance optimization, code translation efficiency, agent-based task solving, and multi-modal code understanding.

208 stars. No commits in the last 6 months.