Flipping Against All Odds: Reducing LLM Coin Flip Bias via Verbalized Rejection Sampling

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Xiao, Tim Z., Zenn, Johannes, Liu, Zhen, Liu, Weiyang, Bamler, Robert, Schölkopf, Bernhard
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866913056766820352
author Xiao, Tim Z.
Zenn, Johannes
Liu, Zhen
Liu, Weiyang
Bamler, Robert
Schölkopf, Bernhard
author_facet Xiao, Tim Z.
Zenn, Johannes
Liu, Zhen
Liu, Weiyang
Bamler, Robert
Schölkopf, Bernhard
contents Large language models (LLMs) can often accurately describe probability distributions using natural language, yet they still struggle to generate faithful samples from them. This mismatch limits their use in tasks requiring reliable stochasticity, such as Monte Carlo methods, agent-based simulations, and randomized decision-making. We investigate this gap between knowledge and sampling in the context of Bernoulli distributions. We introduce Verbalized Rejection Sampling (VRS), a natural-language adaptation of classical rejection sampling that prompts the LLM to reason about and accept or reject proposed samples. Despite relying on the same Bernoulli mechanism internally, VRS substantially reduces sampling bias across models. We provide theoretical analysis showing that, under mild assumptions, VRS improves over direct sampling, with gains attributable to both the algorithm and prompt design. More broadly, our results show how classical probabilistic tools can be verbalized and embedded into LLM workflows to improve reliability, without requiring access to model internals or heavy prompt engineering.
format Preprint
id arxiv_https___arxiv_org_abs_2506_09998
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Flipping Against All Odds: Reducing LLM Coin Flip Bias via Verbalized Rejection Sampling
Xiao, Tim Z.
Zenn, Johannes
Liu, Zhen
Liu, Weiyang
Bamler, Robert
Schölkopf, Bernhard
Machine Learning
Computation and Language
Large language models (LLMs) can often accurately describe probability distributions using natural language, yet they still struggle to generate faithful samples from them. This mismatch limits their use in tasks requiring reliable stochasticity, such as Monte Carlo methods, agent-based simulations, and randomized decision-making. We investigate this gap between knowledge and sampling in the context of Bernoulli distributions. We introduce Verbalized Rejection Sampling (VRS), a natural-language adaptation of classical rejection sampling that prompts the LLM to reason about and accept or reject proposed samples. Despite relying on the same Bernoulli mechanism internally, VRS substantially reduces sampling bias across models. We provide theoretical analysis showing that, under mild assumptions, VRS improves over direct sampling, with gains attributable to both the algorithm and prompt design. More broadly, our results show how classical probabilistic tools can be verbalized and embedded into LLM workflows to improve reliability, without requiring access to model internals or heavy prompt engineering.
title Flipping Against All Odds: Reducing LLM Coin Flip Bias via Verbalized Rejection Sampling
topic Machine Learning
Computation and Language
url https://arxiv.org/abs/2506.09998