The paper evaluates conditional generative surrogates used in high‑energy physics, noting that common 1D marginal divergence metrics can mistakenly reward models that ignore conditioning information. Using the Continuous Ranked Probability Skill Score (CRPSS) as a more robust metric, the authors demonstrate that a simple sampler can outperform a trained Mixture Density Network on all target marginals, and that the advantage diminishes when CRPSS is evaluated within bins of the conditioning variable. They further introduce a data‑driven k‑Nearest‑Neighbor extrapolation to estimate the maximum achievable CRPSS given the conditioning information, finding that both models reach this estimate, suggesting current architectures have extracted the available conditioning information for the evaluated targets.
Conditional generative surrogates in high-energy physics, such as models that predict reconstructed jets from partons, are often evaluated on $1\text{D}$ marginal divergences among other metrics. Take...
arXiv:2605. 12951v2 Announce Type: replace-cross Abstract: We propose Coreset-Induced Conditional Velocity Flow Matching (CCVFM), a generative model that augments hierarchical rectified flow with a data-informed source distribution.
arXiv:2609.23789v1 Announce Type: new Abstract: Modern conditional generative models face significant challenges when learning complex covariate dependencies. While sufficient dimension reduction (SD...
arXiv:2608. 11544v1 Announce Type: cross Abstract: We propose CVaR-penalized Generative Particle Algorithm (CVaR-GPA), a robust, tail-agnostic algorithm for fine-tuning generative models to learn heavy-tailed distributions and capture extreme events, requiring no prior knowledge or estimation of the target's tail characteristics.
Sufficiently Reduced Distributional Regression (SRDR) is a generative approach that merges conditional distribution estimation with nonlinear sufficient dimension reduction (SDR). By framing SDR as a risk minimization problem using strictly proper scoring rules, SRDR jointly learns a dimension reduction map and a generative prediction model through minimization of the energy score, which can be estimated via sampling. The method extends to multi‑environment data and classification, and theoretical results show convergence of estimated conditional distributions in energy distance, implying asymptotic sufficiency. In experiments on CT slice localization, superconductivity data, and digit classification, SRDR recovers low‑dimensional sufficient structure and matches or surpasses state‑of‑the‑art nonlinear SDR methods in representation quality and predictive performance.
arXiv:2606. 04342v1 Announce Type: cross Abstract: Multi-step time series forecasting (MSF) is commonly evaluated using point-wise error metrics such as mean squared error (MSE), implicitly treating the conditional mean as a sufficient target.
arXiv:2605. 08446v3 Announce Type: replace Abstract: Bayesian neural networks are typically trained against the evidence lower bound (ELBO), whose Jensen gap closes only when the variational posterior is exact.
arXiv:2608. 06107v1 Announce Type: new Abstract: Machine learning offers a promising avenue to accelerate physical simulations by replacing computationally expensive traditional Partial Differential Equation (PDE) solvers with fast, differentiable surrogate models.
arXiv:2607. 01171v1 Announce Type: new Abstract: Sample-based generative models are increasingly used for probabilistic forecasting in high-stakes decision settings, yet their training objectives are blind to the decision maker's cost structure.
The paper introduces a new approach to source distribution estimation (SDE) in simulation-based science, addressing limitations of existing methods that rely on a fixed surrogate likelihood. By employing an expectation‑maximization framework, the authors iteratively train an amortized posterior on fresh simulations (E‑step) and refit the source distribution to the posterior’s average (M‑step). Two parameterizations are explored: separate source and posterior flows, and a single shared conditional flow, with experiments on three benchmark tasks showing improved performance over fixed surrogate and iterated baseline methods, notably achieving higher data‑space C2ST scores on the Lotka–Volterra benchmark.
The paper introduces the transport Neural Tangent Kernel (tNTK) to enable active learning for conditional generative models that produce distributional outputs. By decomposing the Wasserstein test risk into mean and shape components, the authors show that mean‑based active learning misses distributional shape, and propose tNTK as a tractable kernel that captures sensitivity of the predicted distribution to model parameters. Using tNTK as a Gaussian process covariance, they greedily select inputs that minimize posterior variance, and demonstrate that this distribution‑aware strategy outperforms mean‑based NTK baselines across several generative model types and datasets.
Conditional generative models have emerged as powerful tools for sampling from target conditional distributions, driving substantial advances across a wide range of scientific and applied domains. As these models proliferate, practitioners often face multiple plausible generators whose performance can vary with the task, data, or input condition.