用球面原生架构提升分子属性预测，更少参数表现更强。
Chem-GMNet: A Sphere-Native Geometric Transformer for Molecular Property Prediction
- 将分子结构映射到球面上，用几何先验替代文本式建模。
- 零预训练下在ESOL任务上达0.938 RMSE，优于预训练ChemBERTa。
- 适合追求高效高精度的分子表征与生成研究者使用。
当前基于SMILES的化学语言模型通过将SMILES视为通用文本，并借助数百万分子的自监督预训练获得优异的MoleculeNet性能。我们提出疑问：当化学领域具有如此丰富的结构先验时，是否应采用领域原生的Transformer，而非依赖规模弥补的通用模型？本文正面回答此问题，提出 extbf{GM-Net}（几何测度网络），其所有模块均为球面原生设计，并实例化为 extbf{Chem-GMNet}。核心组件包括：SH-Embedding（通过Gegenbauer特征映射将令牌表示为$S^{k-1}$上的可学习方向）；DualSKA（每头融合线性时间门控球面流递归与软最大值球面核分支，其持久状态被证明是输入分布的截断多极展开）；以及SH-FFN（球面投影→Gegenbauer提升→矩读出）。在标准DeepChem骨架分割下，与相同规模的ChemBERTa-2基线在chemberta3-faithful协议下对比：(i) 随机初始化的Chem-GMNet在10个MoleculeNet任务中胜出7项，参数量减少约35%；(ii) 在与ChemBERTa-2相同的1000万条SMILES ZINC语料上预训练后，其在8个共享任务中达到或超过公开版本表现（5/7，排除已知ClinTox异常）。$(k,L)$消融实验显示，固定$L=3$，将球面维度从$k=8$增至$k=10$，可使未预训练的ESOL RMSE降至0.938，超越预训练的ChemBERTa-2 MLM-10M。
原文摘要 · Abstract (English)
Modern SMILES-based chemical language models obtain strong MoleculeNet performance by treating SMILES as generic text and compensating with multi-million-molecule self-supervised pretraining. We ask: when a domain carries structural priors as rich as chemistry's, does it warrant a domain-native transformer rather than a generic one rescued by scale? We answer affirmatively with \textbf{GM-Net} (Geometric Measure Network), a transformer family in which every module is replaced by a sphere-native counterpart, and instantiate it as \textbf{Chem-GMNet}. Three blocks follow: SH-Embedding (tokens as learnable directions on $S^{k-1}$ lifted through a Gegenbauer feature map); DualSKA (a per-head fusion of a linear-time gated Sphere-Flow recurrence whose persistent state we prove is the truncated multipole expansion of the input distribution, and a softmax Sphere-Kernel branch over the same Schoenberg-valid kernel); and SH-FFN (sphere projection $\to$ Gegenbauer lift $\to$ moment readout). On canonical DeepChem scaffold splits, against same-shape ChemBERTa-2 baselines under the chemberta3-faithful protocol: (i) random-initialised, Chem-GMNet wins on 7 of 10 MoleculeNet endpoints at $\sim\!35\%$ fewer parameters; (ii) pretrained on the same 10M-SMILES ZINC corpus as ChemBERTa-2 MLM-10M, it matches or beats the public release on 6 of 8 shared endpoints (5/7 excluding a known ClinTox release anomaly). A $(k,L)$ ablation shows that increasing the sphere dimension from $k\!=\!8$ to $k\!=\!10$ at fixed $L\!=\!3$ lowers ESOL RMSE to $0.938$ at scratch, beating pretrained ChemBERTa-2 MLM-10M on this endpoint without any pretraining at all.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。