Guardado en:
Detalles Bibliográficos
Autores principales: Wang, Haojin, Zhu, Zining, Shi, Freda
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:https://arxiv.org/abs/2505.12244
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866915505483284480
author Wang, Haojin
Zhu, Zining
Shi, Freda
author_facet Wang, Haojin
Zhu, Zining
Shi, Freda
contents Autoregressive neural language models (LMs) generate a probability distribution over tokens at each time step given a prompt. In this work, we attempt to systematically understand the probability distributions that LMs can produce, showing that some distributions are significantly harder to elicit than others. Specifically, for any target next-token distribution over the vocabulary, we attempt to find a prompt that induces the LM to output a distribution as close as possible to the target, using either soft or hard gradient-based prompt tuning. We find that (1) in general, distributions with very low or very high entropy are easier to approximate than those with moderate entropy; (2) among distributions with the same entropy, those containing ''outlier tokens'' are easier to approximate; (3) target distributions generated by LMs -- even LMs with different tokenizers -- are easier to approximate than randomly chosen targets. These results offer insights into the expressiveness of LMs and the challenges of using them as probability distribution proposers.
format Preprint
id arxiv_https___arxiv_org_abs_2505_12244
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Distribution Prompting: Understanding the Expressivity of Language Models Through the Next-Token Distributions They Can Produce
Wang, Haojin
Zhu, Zining
Shi, Freda
Computation and Language
Autoregressive neural language models (LMs) generate a probability distribution over tokens at each time step given a prompt. In this work, we attempt to systematically understand the probability distributions that LMs can produce, showing that some distributions are significantly harder to elicit than others. Specifically, for any target next-token distribution over the vocabulary, we attempt to find a prompt that induces the LM to output a distribution as close as possible to the target, using either soft or hard gradient-based prompt tuning. We find that (1) in general, distributions with very low or very high entropy are easier to approximate than those with moderate entropy; (2) among distributions with the same entropy, those containing ''outlier tokens'' are easier to approximate; (3) target distributions generated by LMs -- even LMs with different tokenizers -- are easier to approximate than randomly chosen targets. These results offer insights into the expressiveness of LMs and the challenges of using them as probability distribution proposers.
title Distribution Prompting: Understanding the Expressivity of Language Models Through the Next-Token Distributions They Can Produce
topic Computation and Language
url https://arxiv.org/abs/2505.12244