Big Code Bench

Evaluates LLM code generation across 1,140 realistic programming tasks organized into two splits (Complete for all models, Instruct for chat models) using pass@k statistical metrics that measure the probability at least one of k generated samples passes all test cases. The system generates multiple code samples per task, executes each against embedded test suites, and aggregates results into pass@1, pass@10, pass@100 metrics for comparative model analysis.

Provides a unified interface for generating code samples across heterogeneous LLM providers (OpenAI, Anthropic, Ollama, local models) through a provider-agnostic abstraction that handles API differences, authentication, and response parsing. The system maps provider-specific APIs to a common code generation interface, enabling seamless model swapping without changing benchmark code.

Supports configurable generation parameters (temperature, top_p, max_tokens, n_samples) that control LLM sampling behavior and output diversity. Users can specify different parameter sets per model, enabling exploration of temperature-quality tradeoffs and sample efficiency without code changes.

Executes generated code samples in isolated environments using pluggable backends (local execution with safety limits, E2B sandbox for remote execution, Hugging Face Gradio spaces) that prevent malicious or buggy code from affecting the host system. Each backend enforces resource limits, timeout constraints, and dependency isolation while capturing stdout/stderr and execution results for evaluation.

Pre-processes generated code through a sanitization pipeline that removes unsafe patterns (e.g., file system operations, network calls) and validates Python syntax using AST parsing before execution. The system identifies and flags code that violates safety constraints, preventing execution of malicious or structurally invalid code while maintaining semantic correctness for legitimate implementations.

Manages a curated dataset of 1,140 programming tasks organized into two splits (Complete for all models, Instruct for instruction-tuned models) and two difficulty subsets (full benchmark, hard subset with 148 challenging tasks). Each task includes docstrings, natural language instructions, test cases, and metadata enabling stratified evaluation across model types and difficulty levels.

Aggregates per-task execution results into statistical pass@k metrics that estimate the probability at least one of k generated samples passes all test cases. The system computes pass@1, pass@10, pass@100 from raw execution results, handles edge cases (fewer than k samples generated), and produces leaderboard-formatted output for model comparison.

Exposes four main CLI commands (generate, evaluate, syncheck, inspect) that decompose the benchmark workflow into discrete, composable steps. Users can generate code samples, validate syntax, execute evaluations, and analyze results independently, enabling partial re-runs, debugging, and custom pipeline construction without re-generating all samples.

Persists generated code samples and evaluation results to disk using structured formats (JSONL for samples, JSON for metrics) organized by model, split, backend, and temperature. The system maintains consistent file naming conventions enabling result tracking, comparison, and re-analysis without re-running evaluations.

Provides pre-configured Docker templates (e2b.Dockerfile, e2b.toml) for deploying isolated code execution environments via E2B sandbox service. Templates define base image, dependency installation, resource limits, and timeout configuration, enabling reproducible remote execution without manual environment setup.

Executes generated code against embedded test cases for each task, capturing execution results (pass/fail), stdout/stderr output, execution time, and error traces. The system handles test case isolation, timeout enforcement, and exception handling to produce reliable pass/fail verdicts even when code crashes or hangs.