arXiv：cs.LG· Sam Gijsen, Moritz Seiler, Kerstin Ritter·· 3 小时前
研究揭示条件流匹配模型会幻觉出科学交互效应
Predicting Unseen Scientific Interactions: Conditional Flow Models Hallucinate Non-Additive Effects
AI 导读
研究用生理时间序列模拟器、药物剂量响应数据、组合细胞扰动和分子构象生成，考察流匹配模型对科学交互效应的恢复能力。结果显示模型学习单个效应远早于交互效应，且会在不存在交互的地方幻觉出交互结构，聚合保真度指标可能掩盖这一失败。目标越难识别幻觉越多，输出空间成功恢复的交互在非线性读出下也可能失效。
正文
Abstract:Conditional generative models are increasingly used for scientific prediction beyond the direct training support. Yet scientific questions often concern whether one intervention changes the effect of another. Such interactions can be of primary interest while explaining little of the total variation, allowing models to appear accurate overall without recovering the relevant scientific effect. It remains unclear how these models learn interactions and when these successfully extrapolate to unseen conditions. We therefore study interaction recovery in flow-matching models using a physiological time-series simulator, drug dose-response data, combinatorial cell perturbations, and molecular conformer generations, where we separate recovery from hallucinated interaction structure. In the controlled setting, individual effects are learned much earlier than interactions and models hallucinate interactions where none exist. Across settings, we find qualitatively different generalization problems, with successful prediction depending on whether condition structure continues from observed to held-out targets. When it does, interactions are recoverable. If not, models can generate plausible responses while missing the interaction. We furthermore find that hallucinations increase as targets become less identifiable, and that successful interaction recovery in output space can fail under nonlinear readouts. Finally, we show that aggregate fidelity metrics can conceal whether models learn and extrapolate interactions.
| Comments: | Accepted as an oral at NeurIPS 2026 Workshop: AI Data Readiness for Scientific Discovery (AIDaR). Code: this https URL |
| Subjects: | Machine Learning (cs.LG); Quantitative Methods (q-bio.QM) |
| Cite as: | arXiv:2610.20288 [cs.LG] |
| (or arXiv:2610.20288v1 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2610.20288 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Sam Gijsen [view email]
[v1]
Wed, 14 Oct 2026 14:42:12 UTC (147 KB)
来源：arXiv：cs.LG · arxiv.org