Overview

Rare-event estimation seeks reliable probabilities for outcomes that occur too infrequently for ordinary Monte Carlo sampling, such as system failures, climate extremes, unsafe language-model outputs, or molecular transitions. Common approaches include importance sampling, subset simulation and multilevel splitting, cross-entropy methods, and sequential Monte Carlo; newer methods use learned or generative proposals to steer samples toward relevant outcomes. These techniques can reduce the number of simulations needed, but their usefulness depends on finding the important event regions while keeping estimates unbiased and variance manageable. A continuing challenge is validating estimates when events are extremely rare and results depend on model calibration, proposal quality, or the distribution of inputs being evaluated.

Latest papers

CardsList

34 papers7 papers in the last four weeks, down 22% on the four weeks before. 0.1% of all new papers.

  1. Differentiable free energy surface: a variational approach to directly observing rare events using generative deep-learning models

    Apr 10, 2026Shuo-Hui Li, Chen Chen, Yao-Wen Zhang +1Free Energy CalculationRare-Event Estimation

  2. Reasoning-Enhanced Rare-Event Prediction with Balanced Outcome Correction

    Jan 23, 2026Vitaly Bulgakov, Alexander TurchinClass-Imbalanced LearningRare-Event Prediction