VEHBench - Policy-Conditioned Selection: leaderboard
Metric: P4 Kendall tau-b (-1 to 1): rank agreement between the model ranking of a feasible candidate pool and the oracle ranking under an explicit deployment policy; LLM-assisted vibration energy harvester design against a physics oracle; one complete run per model. Source: arxiv.org. Saturation forecast: Estimated already saturated. 12 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.4 | 0.89 |
| 2 | Qwen 3.6 Plus | 0.88 |
| 3 | DeepSeek V3 | 0.86 |
| 4 | MiMo-V2.5-Pro (Non-reasoning) | 0.84 |
| 5 | Claude Sonnet 4.6 | 0.84 |
| 6 | Hy3-preview (Reasoning) | 0.84 |
| 7 | Qwen 3 Max | 0.84 |
| 8 | DeepSeek R1 | 0.83 |
| 9 | Gemini 3.1 Pro (Preview) | 0.82 |
| 10 | DeepSeek V4 Pro (Non-reasoning) | 0.79 |
| 11 | O4 Mini | 0.78 |
| 12 | Llama 3.3 70B | 0.71 |
Interactive version: theaggregate.ai/benchmark?slug=vehbench-policy-conditioned-selection · How It Works · Data refreshed daily, snapshot 2026-09-29.