Discovering and Learning Probabilistic Models of Black-Box AI Capabilities

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Bramblett, Daniel, Karia, Rushang, Ciotinga, Adrian, Suresh, Ruthvick, Verma, Pulkit, Choi, YooJung, Srivastava, Siddharth
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866908724574027776
author Bramblett, Daniel
Karia, Rushang
Ciotinga, Adrian
Suresh, Ruthvick
Verma, Pulkit
Choi, YooJung
Srivastava, Siddharth
author_facet Bramblett, Daniel
Karia, Rushang
Ciotinga, Adrian
Suresh, Ruthvick
Verma, Pulkit
Choi, YooJung
Srivastava, Siddharth
contents Black-box AI (BBAI) systems such as foundational models are increasingly being used for sequential decision making. To ensure that such systems are safe to operate and deploy, it is imperative to develop efficient methods that can provide a sound and interpretable representation of the BBAI's capabilities. This paper shows that PDDL-style representations can be used to efficiently learn and model an input BBAI's planning capabilities. It uses the Monte-Carlo tree search paradigm to systematically create test tasks, acquire data, and prune the hypothesis space of possible symbolic models. Learned models describe a BBAI's capabilities, the conditions under which they can be executed, and the possible outcomes of executing them along with their associated probabilities. Theoretical results show soundness, completeness and convergence of the learned models. Empirical results with multiple BBAI systems illustrate the scope, efficiency, and accuracy of the presented methods.
format Preprint
id arxiv_https___arxiv_org_abs_2512_16733
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Discovering and Learning Probabilistic Models of Black-Box AI Capabilities
Bramblett, Daniel
Karia, Rushang
Ciotinga, Adrian
Suresh, Ruthvick
Verma, Pulkit
Choi, YooJung
Srivastava, Siddharth
Artificial Intelligence
Black-box AI (BBAI) systems such as foundational models are increasingly being used for sequential decision making. To ensure that such systems are safe to operate and deploy, it is imperative to develop efficient methods that can provide a sound and interpretable representation of the BBAI's capabilities. This paper shows that PDDL-style representations can be used to efficiently learn and model an input BBAI's planning capabilities. It uses the Monte-Carlo tree search paradigm to systematically create test tasks, acquire data, and prune the hypothesis space of possible symbolic models. Learned models describe a BBAI's capabilities, the conditions under which they can be executed, and the possible outcomes of executing them along with their associated probabilities. Theoretical results show soundness, completeness and convergence of the learned models. Empirical results with multiple BBAI systems illustrate the scope, efficiency, and accuracy of the presented methods.
title Discovering and Learning Probabilistic Models of Black-Box AI Capabilities
topic Artificial Intelligence
url https://arxiv.org/abs/2512.16733