Latent Veracity Inference for Identifying Errors in Stepwise Reasoning

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Kim, Minsu, Falet, Jean-Pierre, Richardson, Oliver E., Chen, Xiaoyin, Jain, Moksh, Ahn, Sungjin, Ahn, Sungsoo, Bengio, Yoshua
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866908836931043328
author Kim, Minsu
Falet, Jean-Pierre
Richardson, Oliver E.
Chen, Xiaoyin
Jain, Moksh
Ahn, Sungjin
Ahn, Sungsoo
Bengio, Yoshua
author_facet Kim, Minsu
Falet, Jean-Pierre
Richardson, Oliver E.
Chen, Xiaoyin
Jain, Moksh
Ahn, Sungjin
Ahn, Sungsoo
Bengio, Yoshua
contents Chain-of-Thought (CoT) reasoning has advanced the capabilities and transparency of language models (LMs); however, reasoning chains can contain inaccurate statements that reduce performance and trustworthiness. To address this, we propose to augment each reasoning step in a CoT with a latent veracity (or correctness) variable. To efficiently explore this expanded space, we introduce Veracity Search (VS), a discrete search algorithm over veracity assignments. It performs otherwise intractable inference in the posterior distribution over latent veracity values by leveraging the LM's joint likelihood over veracity and the final answer as a proxy reward. This efficient inference-time verification method facilitates supervised fine-tuning of an Amortized Veracity Inference (AVI) machine by providing pseudo-labels for veracity. AVI generalizes VS, enabling accurate zero-shot veracity inference in novel contexts. Empirical results demonstrate that VS reliably identifies errors in logical (ProntoQA), mathematical (GSM8K), and commonsense (CommonsenseQA) reasoning benchmarks, with AVI achieving comparable zero-shot accuracy. Finally, we demonstrate the utility of latent veracity inference for providing feedback during self-correction and self-improvement.
format Preprint
id arxiv_https___arxiv_org_abs_2505_11824
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Latent Veracity Inference for Identifying Errors in Stepwise Reasoning
Kim, Minsu
Falet, Jean-Pierre
Richardson, Oliver E.
Chen, Xiaoyin
Jain, Moksh
Ahn, Sungjin
Ahn, Sungsoo
Bengio, Yoshua
Machine Learning
Artificial Intelligence
Chain-of-Thought (CoT) reasoning has advanced the capabilities and transparency of language models (LMs); however, reasoning chains can contain inaccurate statements that reduce performance and trustworthiness. To address this, we propose to augment each reasoning step in a CoT with a latent veracity (or correctness) variable. To efficiently explore this expanded space, we introduce Veracity Search (VS), a discrete search algorithm over veracity assignments. It performs otherwise intractable inference in the posterior distribution over latent veracity values by leveraging the LM's joint likelihood over veracity and the final answer as a proxy reward. This efficient inference-time verification method facilitates supervised fine-tuning of an Amortized Veracity Inference (AVI) machine by providing pseudo-labels for veracity. AVI generalizes VS, enabling accurate zero-shot veracity inference in novel contexts. Empirical results demonstrate that VS reliably identifies errors in logical (ProntoQA), mathematical (GSM8K), and commonsense (CommonsenseQA) reasoning benchmarks, with AVI achieving comparable zero-shot accuracy. Finally, we demonstrate the utility of latent veracity inference for providing feedback during self-correction and self-improvement.
title Latent Veracity Inference for Identifying Errors in Stepwise Reasoning
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2505.11824