CADGenBench: Inputs for AI-Generated 3D Mechanical Parts Benchmark

CADGenBench provides public inputs for a benchmark measuring AI systems' ability to produce correct 3D mechanical parts as STEP files. The dataset, created by HuggingAI4Engineering, contains only the task inputs, with ground truth withheld in a separate private repository to maintain evaluation integrity. It was last updated on June 8, 2026.

Use Cases

• Benchmarking AI model performance on 3D CAD generation based on the described task inputs
• Developing AI systems for mechanical part design based on the benchmark's focus on STEP files
• Evaluating the correctness of AI-generated 3D geometries for engineering applications
• Comparing different AI approaches for 3D mechanical part creation

Strengths

• Designed for a specific, structured benchmark (CADGenBench) with a clear evaluation goal
• Maintains a single source of truth for evaluation by withholding ground truth in a separate repository
• Last updated on 2026-06-08, suggesting recent maintenance

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download
• Column-level documentation is absent; field semantics must be inferred after download
• Row count is unknown, which may limit suitability assessment

Provenance

Source

HuggingAI4Engineering

Collection Method

Likely created as part of a benchmark initiative for AI in engineering.

Freshness

Last updated 2026-06-08 13:46:35; freshness should be verified