Unverified50% confidenceFactExact time
Benchmark datasets in materials science are typically limited in scale (tens to hundreds of samples) and differ in distribution from the large-scale synthetic data that TFMs encounter during pretraining; the heterogeneity of materials data may fall outside the pretraining distribution range.
1
Sources
50%
Confidence
Long-term
Relevance
11/10/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedHumanEval has a limited sample size and has been extensively covered by training data, leading to suspicions of test set leakage in some models69% similarUnverifiedTraining on a dataset with an overly homogeneous object type causes the model's performance to drop sharply when faced with objects of other materials or forms, which is the domain shift problem.66% similarUnverified预处理泄漏在小数据集上尤为显著,当训练集仅有数百张图像时加入测试集样本会明显改变均值和标准差估计65% similarUnverifiedStatistical methods rely on signals such as TF-IDF and co-occurrence frequency, require no training data, and remain strong baselines in low-resource settings.64% similarUnverifiedDomain adaptation tasks typically require hundreds to thousands of high-quality samples, and mixing in large amounts of low-quality data will instead degrade model performance.63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1155146API
curl https://kongchang.com/api/v1/knowledge/claims/1155146MCP
get_claim(id=1155146)