AI

GenPFN: An In-Context Tabular Foundation Model Approach for Augmenting Small Datasets in Engineering

Read full article at source ↗

Based on a feed excerpt; full article content has not been acquired. Source evidence retrieved 13/11/2026, 07:48:24 UTC.

Briefings

GenPFN is a training-free candidate generator that uses a pretrained TabPFN classifier to augment small engineering datasets. Evaluated on fourteen problems, it outperformed adversarial and diffusion generators in class satisfaction and realism distance, while SMOTE showed higher coverage distance. The hull-stern case study achieved 99.3% propeller-compatible candidates versus 1.4% for random sampling, reducing validation MSE by 62.1% compared to context-only training.

Key points

• GenPFN uses a pretrained TabPFN classifier with fixed weights to generate candidates without training.
• Evaluated on fourteen problems, GenPFN achieved higher class satisfaction and lower realism distance than adversarial and diffusion generators.

Context

Engineering design often faces costly simulations and experiments limiting available data. This paper proposes GenPFN to augment limited datasets using a pretrained tabular classifier.

Limitations

Evaluation spans only fourteen selected problems, including a two-parameter benchmark and a 44-parameter hull problem. Results are from paired comparisons and regression fits; specific causal claims about the mechanism remain unverified.

Generated by Bittrees-hosted qwen35-2b sha256:aaf42c8b7c3c on 13/11/2026, 08:42:24 UTC. Based on publisher-provided feed evidence; may not cover the full original article. Check the source for details.

View archived briefing on IPFS
bafybeihgsbkbllsqvtgoxd5mj6conzm3mtprxbq2gnbtzq75d22d4gsm3u