Estimating Rare Events in Language Models with Proper Evaluation

• URL: http://arxiv.org/abs/2607.18454v1
• Date: Mon, 20 Jul 2026 19:07:43 GMT
• Title: Estimating Rare Events in Language Models with Proper Evaluation
• Abstract summary: We introduce Gradient Activation Adaptive Multi-Level Splitting (GA-AMLS)<n>GA-AMLS adapts rare-event Monte Carlo methods to the continuous activation space of language models.<n>We also propose the Shifted-Power Bregman (SPB) Loss, a proper scoring rule that remains finite for zero-estimates.
• Score: 16.1901795614153
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Quantifying the risk of rare failures in language models, such as those triggered by adversarial distribution shifts or very large-scale deployments, requires estimating probabilities far too small for random sampling. While recent work has formalized Low Probability Estimation, existing pipelines remain fragile in the rarest regimes: estimators can suffer zero-estimate collapse or systematic bias, and standard evaluation losses can become unstable or poorly matched to asymmetric safety costs. In this work, we introduce Gradient Activation Adaptive Multi-Level Splitting (GA-AMLS), which adapts rare-event Monte Carlo methods to the continuous activation space of language models. Specifically, GA-AMLS uses a gradient-based MCMC kernel to navigate activation space, eliminating the zero-estimate collapse of input-space search and replacing the independence assumptions of prior activation-space estimators with conditional sampling under an explicit, heavier-tailed activation prior. We also propose the Shifted-Power Bregman (SPB) Loss, a proper scoring rule that remains finite for zero-estimates and offers tunable asymmetry between underestimation and overestimation penalties. Experiments on small transformer models reveal a bias-variance tradeoff: GA-AMLS achieves the lowest loss under symmetric evaluation, reducing average log-space squared error relative to the strongest baseline across model sizes, while methods with overestimation bias prevail under asymmetric penalties. Our findings highlight that estimator choice should be matched to deployment context. More broadly, our work establishes activation space as a tractable domain for rare-event estimation in language models, circumventing the brittleness of discrete input-space search.

Related papers

• Mind the Residual Gap: Probabilistic Downscaling under Real-World Bias [11.400279490725042]
  Probabilistic downscaling is a central challenge to atmospheric science, climate modeling, and other multiscale physical systems.<n>A widely used paradigm decomposes the problem into a deterministic mean predictor followed by a residual generator.<n>We show that the root cause is residual target miscalibration, the residual distribution induced during training differs systematically from the one required at test time due to downscaling bias.
  arXiv  Detail & Related papers  (2026-06-29T18:47:34Z)
• Using Large Language Models as Low-Cost Statistical Estimators for Human-Response Data [0.282023281338763]
  We show that pretrained large language models induce risk-equivalent estimators of conditional expectations under squared loss.<n>We provide finite-sample concentration bounds and a calibration protocol with explicit decision rules.
  arXiv  Detail & Related papers  (2026-06-29T14:34:28Z)
• How Useful is Causal Invariance for Domain Adaptation in Finite-Sample Settings? [58.740078141879984]
  Machine learning models often degrade when they are deployed on a target distribution that differs from the source distributions they were trained on.<n>Recent work in causality-based domain generalization has shown how shared causal structure can induce invariant predictors.<n>In this paper, we study when full or partial causal knowledge can provably improve supervised domain adaptation.
  arXiv  Detail & Related papers  (2026-06-10T21:07:49Z)
• VAE-Inf: A statistically interpretable generative paradigm for imbalanced classification [8.677199689027772]
  We propose a two-stage framework to bridge the gap between generative modeling and discriminative classification.<n>For inference, we introduce a projection-based score that admits a natural hypothesis testing interpretation.<n>Experiments on diverse real-world benchmarks demonstrate that our framework achieves competitive performance against other approaches.
  arXiv  Detail & Related papers  (2026-04-28T07:50:56Z)
• Causal Inference as Distribution Adaptation: Optimizing ATE Risk under Propensity Uncertainty [0.0]
  We reframing ATE estimation as a textitdomain adaptation problem under distribution shift.<n>We propose the textbfJoint Robust Estimator (JRE) to train outcome models jointly.
  arXiv  Detail & Related papers  (2025-12-19T21:40:46Z)
• Relaxed Quantile Regression: Prediction Intervals for Asymmetric Noise [51.87307904567702]
  Quantile regression is a leading approach for obtaining such intervals via the empirical estimation of quantiles in the distribution of outputs.<n>We propose Relaxed Quantile Regression (RQR), a direct alternative to quantile regression based interval construction that removes this arbitrary constraint.<n>We demonstrate that this added flexibility results in intervals with an improvement in desirable qualities.
  arXiv  Detail & Related papers  (2024-06-05T13:36:38Z)
• The Decaying Missing-at-Random Framework: Model Doubly Robust Causal Inference with Partially Labeled Data [8.916614661563893]
  We introduce a missing-at-random (decaying MAR) framework and associated approaches for doubly robust causal inference.<n>This simultaneously addresses selection bias in the labeling mechanism and the extreme imbalance between labeled and unlabeled groups.<n>To ensure robust causal conclusions, we propose a bias-reduced SS estimator for the average treatment effect.
  arXiv  Detail & Related papers  (2023-05-22T07:37:12Z)
• Distributionally Robust Models with Parametric Likelihood Ratios [123.05074253513935]
  Three simple ideas allow us to train models with DRO using a broader class of parametric likelihood ratios. We find that models trained with the resulting parametric adversaries are consistently more robust to subpopulation shifts when compared to other DRO approaches.
  arXiv  Detail & Related papers  (2022-04-13T12:43:12Z)
• Evaluating Distributional Distortion in Neural Language Modeling [81.83408583979745]
  A heavy-tail of rare events accounts for a significant amount of the total probability mass of distributions in language. Standard language modeling metrics such as perplexity quantify the performance of language models (LM) in aggregate. We develop a controlled evaluation scheme which uses generative models trained on natural data as artificial languages.
  arXiv  Detail & Related papers  (2022-03-24T01:09:46Z)
• Near-optimal inference in adaptive linear regression [60.08422051718195]
  Even simple methods like least squares can exhibit non-normal behavior when data is collected in an adaptive manner. We propose a family of online debiasing estimators to correct these distributional anomalies in at least squares estimation. We demonstrate the usefulness of our theory via applications to multi-armed bandit, autoregressive time series estimation, and active learning with exploration.
  arXiv  Detail & Related papers  (2021-07-05T21:05:11Z)
• Amortized Conditional Normalized Maximum Likelihood: Reliable Out of Distribution Uncertainty Estimation [99.92568326314667]
  We propose the amortized conditional normalized maximum likelihood (ACNML) method as a scalable general-purpose approach for uncertainty estimation. Our algorithm builds on the conditional normalized maximum likelihood (CNML) coding scheme, which has minimax optimal properties according to the minimum description length principle. We demonstrate that ACNML compares favorably to a number of prior techniques for uncertainty estimation in terms of calibration on out-of-distribution inputs.
  arXiv  Detail & Related papers  (2020-11-05T08:04:34Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.