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

Indonesian

Model generatif kondisional semakin banyak digunakan untuk prediksi ilmiah di luar dukungan pelatihan langsung. Namun pertanyaan ilmiah sering mengkhawatirkan apakah satu intervensi mengubah efek dari yang lain. Interaksi semacam itu dapat menjadi minat utama sementara menjelaskan sedikit dari variasi total, memungkinkan model untuk muncul akurat secara keseluruhan tanpa memulihkan efek ilmiah yang relevan. Masih tidak jelas bagaimana model ini belajar interaksi hallucinated dan ketika ini berhasil ekstrapolasi ke kondisi yang tidak terlihat. Oleh karena itu kita mempelajari pemulihan interaksi dalam model flow-matching menggunakan simulator serangkaian waktu fisiologis, data interaksi dosis obat, gangguan sel kombinasi, dan generasi konformer molekuler, di mana kita memisahkan pemulihan dari struktur interaksi hallucinated. Dalam pengaturan terkontrol, efek individu dipelajari jauh lebih awal daripada interaksi dan model hallucinate di mana tidak ada

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