Sign in to view source links and access this dataset
Description
18,000 evaluation samples across six benchmark families derived from DeepCAD, Fusion 360, ABC, MCB, and Objaverse. CADBench is a unified multimodal benchmark for AI-assisted CAD program generation, introduced by DeCoDELab. The dataset was last updated on June 10, 2026.
Use Cases
Benchmarking AI models for editable 3D reconstruction based on the described evaluation samples.
Measuring progress in multimodal CAD understanding based on the benchmark's design.
Training AI systems for CAD program generation based on the unified multimodal benchmark.
Strengths
18,000 evaluation samples provide a substantial testbed.
Spans six distinct benchmark families, suggesting diversity in tasks.
Derived from multiple established sources including DeepCAD, Fusion 360, and Objaverse.
Limitations
Column-level documentation is absent; field semantics must be inferred after download.
Row count is unknown, which may limit suitability assessment.
Description metadata is limited; actual data quality requires manual inspection after download.
Provenance
Source
DeCoDELab
Freshness
Last updated 2026-06-10 02:01:54; freshness should be verified.
License is unknown; users should verify terms before use.