Zero-Shot Verification-guided Chain of Thoughts
Jishnu Ray Chowdhury, Cornelia Caragea
TL;DR
The paper tackles zero-shot verification for chain-of-thought reasoning in large language models by introducing COT STEP for structured, stepwise thought decomposition and two zero-shot verifiers. It evaluates across open models (SOLAR and Phi3) on diverse datasets (GSM8K, GSM-HARD, AQuA, StrategyQA, CSQA), comparing against existing zero-shot prompts. Key findings show that COT STEP delivers competitive performance and enables automatic parsing of reasoning steps, while zero-shot verifier scores provide limited gains for self-consistency but can aid in stepwise greedy search in certain settings. The work advances zero-shot reasoning interfaces and lays groundwork for robust verification-augmented prompting in multi-domain tasks.
Abstract
Previous works have demonstrated the effectiveness of Chain-of-Thought (COT) prompts and verifiers in guiding Large Language Models (LLMs) through the space of reasoning. However, most such studies either use a fine-tuned verifier or rely on manually handcrafted few-shot examples. In contrast, in this paper, we focus on LLM-based self-verification of self-generated reasoning steps via COT prompts in a completely zero-shot regime. To explore this setting, we design a new zero-shot prompt, which we call COT STEP, to aid zero-shot decomposition of reasoning steps and design two new zero-shot prompts for LLM-based verifiers. We evaluate the verifiers' ability to classify the correctness of reasoning chains and explore different ways to use verifier scores in guiding reasoning for various mathematical and commonsense reasoning tasks with different LLMs.