Newest papers

Predicting Unseen Scientific Interactions: Conditional Flow Models Hallucinate Non-Additive Effects

Sam Christian Jan Gijsen, Moritz Vinzent Seiler, Kerstin Ritter · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.20288 · cited by 0 works (OpenAlex) · open-access PDF

German

Ob bedingte generative Modelle zunehmend für die wissenschaftliche Vorhersage über die direkte Schulungsunterstützung hinaus verwendet werden. Doch wissenschaftliche Fragen betreffen oft, ob eine Intervention die Wirkung einer anderen verändert. Solche Interaktionen können von vorrangigem Interesse sein, während sie wenig der Gesamtvarianz erklären, so dass Modelle insgesamt präzise erscheinen, ohne den relevanten wissenschaftlichen Effekt wiederherzustellen. Es bleibt unklar, wie diese Modelle Interaktionen lernen und wenn diese erfolgreich auf unsichtbare Bedingungen extrapolieren. Daher studieren wir Interaktionswiederherstellung in Fluss-Matching-Modellen mit einem physiologischen Zeit-Serie-Simulator, Medikamentendosis-Interaktionsdaten, kombinatorische Zellstörungen und molekulare Konformer-Generationen, wo

Machine translation (Meta M2M100); the English original follows.

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.

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: Hallucinating, Generalization, Artificial intelligence, Overfitting, Computer science, Machine learning

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange