Consensus Sampling for Safer Generative AI

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Kalai, Adam Tauman, Kalai, Yael Tauman, Zamir, Or
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866917476021829632
author Kalai, Adam Tauman
Kalai, Yael Tauman
Zamir, Or
author_facet Kalai, Adam Tauman
Kalai, Yael Tauman
Zamir, Or
contents Motivated by undetectable risks in generative AI, we study a general robust aggregation problem: how to aggregate several probability distributions to boost safety. We present consensus sampling, a black-box algorithm that, given k distributions, has risk competitive with the average risk of the safest $s$ while abstaining when there is insufficient agreement. This yields an architecture-agnostic approach to generative-model safety when the distributions are induced by models that can sample and evaluate output probabilities. We formalize the guarantee through R-robustness, which also bounds information leakage and adversarial influence. Inspired by robust statistics and the provable copyright protection algorithm of Vyas et al (2023), we show that while a standard mixture is vulnerable to one unsafe constituent, a pointwise-median construction provides robust intuition, and our efficient sampler is Pareto-optimal for the tradeoff between worst-case risk and abstention. Experiments on synthetic distributions and image generation illustrate the general mechanism and its motivating safety application. The method requires overlap among safe distributions, but it provides a model-agnostic way to inherit guarantees from an unknown reliable subset.
format Preprint
id arxiv_https___arxiv_org_abs_2511_09493
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Consensus Sampling for Safer Generative AI
Kalai, Adam Tauman
Kalai, Yael Tauman
Zamir, Or
Artificial Intelligence
Machine Learning
Motivated by undetectable risks in generative AI, we study a general robust aggregation problem: how to aggregate several probability distributions to boost safety. We present consensus sampling, a black-box algorithm that, given k distributions, has risk competitive with the average risk of the safest $s$ while abstaining when there is insufficient agreement. This yields an architecture-agnostic approach to generative-model safety when the distributions are induced by models that can sample and evaluate output probabilities. We formalize the guarantee through R-robustness, which also bounds information leakage and adversarial influence. Inspired by robust statistics and the provable copyright protection algorithm of Vyas et al (2023), we show that while a standard mixture is vulnerable to one unsafe constituent, a pointwise-median construction provides robust intuition, and our efficient sampler is Pareto-optimal for the tradeoff between worst-case risk and abstention. Experiments on synthetic distributions and image generation illustrate the general mechanism and its motivating safety application. The method requires overlap among safe distributions, but it provides a model-agnostic way to inherit guarantees from an unknown reliable subset.
title Consensus Sampling for Safer Generative AI
topic Artificial Intelligence
Machine Learning
url https://arxiv.org/abs/2511.09493