EngiWorld: What Can Frontier Agents Deliver in Professional Engineering Environments?
The strongest tested agent scored 44.3 on EngiWorld, with multi-software tasks succeeding only 3.6% of the time.
The paper introduces EngiWorld, a benchmark of 1,301 expert-curated engineering tasks across CAD, CAE, CAM, BIM, EDA, and 3D visualization. It spans 26 professional software platforms and includes both GUI and CLI workflows. The authors evaluate artifacts with domain verifiers that check geometry, physics, rule compliance, and intermediate outputs. Their result is that frontier agents still fall well short of reliable end-to-end professional engineering automation. HF Daily Papers' note
The paper introduces EngiWorld, a benchmark of 1,301 expert-curated engineering tasks across CAD, CAE, CAM, BIM, EDA, and 3D visualization. It spans 26 professional software platforms and includes both GUI and CLI workflows. The authors evaluate artifacts with domain verifiers that check geometry, physics, rule compliance, and intermediate outputs. Their result is that frontier agents still fall well short of reliable end-to-end professional engineering automation. HF Daily Papers' note
score 6