VEHBench: A Stage-Local Diagnostic Benchmark for LLM-Assisted Vibration Energy Harvester Design
VEHBench tests LLMs by design stage, not just by whether a final engineering artifact works.
The paper introduces 763 literature-grounded vibration energy harvester tasks scored by an analytical physical oracle. It evaluates four roles: specification triage, verifier-guided search, corrupted-state recovery, and policy-conditioned selection. The authors report that model performance is stage-dependent, with no single model dominating the whole workflow. Source: ArXiv · AI/CL/LG's note
The paper introduces 763 literature-grounded vibration energy harvester tasks scored by an analytical physical oracle. It evaluates four roles: specification triage, verifier-guided search, corrupted-state recovery, and policy-conditioned selection. The authors report that model performance is stage-dependent, with no single model dominating the whole workflow. Source: ArXiv · AI/CL/LG's note
score 4