A Quiet Failure in Calibrated Virtual Screening: Marginal Conformal Prediction Under-Covers the Minority Class, and a Class-Conditional Fix Recovers It
arXiv:2607.06605v1 Announce Type: new Abstract: Conformal prediction is being adopted in drug discovery to put an honest number on model reliability: pick an error rate alpha, and the method returns prediction sets containing the true label with probability at least 1 - alpha. We show this guarantee can be dangerous on imbalanced datasets. Across four datasets, standard (marginal) conformal prediction hits its global 90% coverage target while leaving the minority class badly exposed: realized minority coverage falls to 64.8% on blood-brain-barrier penetration and to 4.2% on clinical-trial toxi
The increasing adoption of AI in high-stakes fields like drug discovery necessitates robust reliability measures, exposing nuanced limitations in current methods like conformal prediction, particularly with imbalanced data.
This highlights a critical flaw in a widely used method for AI reliability (conformal prediction) in drug discovery, potentially leading to dangerous under-coverage of minority classes and undermining trust in AI-driven medical applications.
The understanding of conformal prediction's safety guarantees changes, requiring class-conditional fixes, moving beyond global coverage metrics to ensure equitable reliability across all data subsets, especially in imbalanced datasets.
- · AI safety researchers
- · Drug discovery companies adopting robust AI
- · Patients with rare diseases
- · Developers relying solely on marginal conformal prediction
- · Drug discovery pipelines with unaddressed data imbalance
Increased scrutiny and demand for class-conditional reliability metrics in AI models, especially in critical applications.
Development of new AI reliability frameworks and tools that explicitly account for data imbalance and minority class performance.
Heightened regulatory requirements for AI in drug discovery to demonstrate equitable performance across diverse patient populations.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.LG