Gradio

BikeBench Leaderboard

Training filters

Conditional?

Masked?

Grad-free?

Max eval count

Reproducibility filters

Has GitHub link?

Has paper link?

Admin-verified?

Performance filters

Min Design Quality (HV)

Max Constraint Violation

Max Sim. to Data (MMD)

Min Novelty

Min Binary Validity

Max Diversity (DPP)

Filter ID

Filter ID (copy/paste to share or load)

Leaderboard

Leaderboard

GPT-5 (Unconditional Non-Interactive Prompt)	0.008653197	1.3013	0.147113495	7.096608908

Leaderboard

CTGAN (Rejection Sampled)	0.00865	0.00000	0.14711	7.09661
CTGAN (Rejection Sampled)	0.01479	0.31800	0.08214	11.92779
CTGAN (Rejection Sampled)	0.00792	1.04600	0.12330	9.94108
CTGAN (Rejection Sampled)	0.01092	1.30130	0.09824	11.65046
DDPM (Optimization-Augmented)	0.01678	0.00000	0.07742	12.55460
Guided DDPM (Optimization-Augmented)	0.01947	0.00000	0.03742	13.02612
Guided DDPM (Optimization-Augmented)	0.01874	0.33940	0.03712	12.97344
C-DDPM (Optimization-Augmented)	0.01645	0.04160	0.08378	12.81821
DDPM (Optimization-Augmented)	0.00870	1.54140	0.10073	11.06186
C-DDPM (Optimization-Augmented)	0.00070	2.49540	0.20146	12.92903
NSGA-II	0.01018	0.00000	0.62165	11.88004
GAN (Optimization-Augmented)	0.01366	0.00000	0.30752	14.43946
C-GAN (Optimization-Augmented)	0.01086	2.57920	0.07860	14.03852
GAN (Optimization-Augmented)	0.01277	0.50520	0.19558	14.02043
C-GAN (Optimization-Augmented)	0.01113	2.38980	0.10480	14.11920
NSGA-II	0.00105	3.56600	0.60640	11.59283
Gradient Descent	0.01642	0.09190	0.26579	12.03611
Gradient Descent	0.01609	0.68330	0.27900	12.08588
TVAE (Rejection Sampled)	0.01602	0.00000	0.16319	13.55419
TVAE (Rejection Sampled)	0.01611	0.05300	0.15332	13.55061
TVAE (Rejection Sampled)	0.01503	0.61610	0.16406	13.54888
TVAE (Rejection Sampled)	0.01507	0.42790	0.15367	13.54291
VAE (Optimization-Augmented)	0.01811	0.00000	0.30092	14.95996
C-VAE (Optimization-Augmented)	0.01863	0.00140	0.34759	14.97072
VAE (Optimization-Augmented)	0.01425	0.97130	0.34055	14.99489
C-VAE (Optimization-Augmented)	0.01572	1.06150	0.28378	14.92151
Random Sample from Dataset (Baseline)	0.00346	2.69880	0.00506	12.72189
Rejection Sample from Dataset (Baseline)	0.01514	0.00000	0.08142	12.27817
Rejection Sample from Dataset (Baseline)	0.01514	0.11160	0.07884	12.43054
Rejection Sample from Dataset (Baseline)	0.01514	0.12320	0.08181	12.41179
Rejection Sample from Dataset (Baseline)	0.01514	0.21570	0.07644	12.51054
GPT-5 (Unconditional Non-Interactive Prompt)	0.00857	0.44770	0.36303	14.07629
GPT-5 (Conditional Non-Interactive Prompt)	0.00491	1.28770	0.63126	14.69556

Project links:
• BikeBench GitHub repository
• BikeBench paper (arXiv)

Citation (BibTeX):

@inproceedings{regenwetter2025bike,
title={BikeBench: A Bicycle Design Benchmark for Generative Models with Objectives and Constraints},
author=Regenwetter, Lyle and Obaideh, Yazan Abu and Chiotti, Fabien and Lykourentzou, Ioanna and Ahmed, Faez},
journal={Advances in Neural Information Processing Systems},
year={2025}
}