Recursive Think-Answer Process for LLMs and VLMs

TL;DR

This paper introduces R-TAP, a recursive reasoning method for LLMs and VLMs that iteratively improves answer accuracy by assessing confidence and reducing errors during inference.

Contribution

The paper proposes R-TAP, an iterative reasoning framework with confidence evaluation that enhances accuracy and stability over traditional single-pass models.

Findings

R-TAP outperforms single-pass methods in accuracy for LLMs and VLMs.

Models with R-TAP show fewer self-reflective errors like "Oops!".

R-TAP leads to more stable and faster inference processes.

Abstract

Think-Answer reasoners such as DeepSeek-R1 have made notable progress by leveraging interpretable internal reasoning. However, despite the frequent presence of self-reflective cues like "Oops!", they remain vulnerable to output errors during single-pass inference. To address this limitation, we propose an efficient Recursive Think-Answer Process (R-TAP) that enables models to engage in iterative reasoning cycles and generate more accurate answers, going beyond conventional single-pass approaches. Central to this approach is a confidence generator that evaluates the certainty of model responses and guides subsequent improvements. By incorporating two complementary rewards-Recursively Confidence Increase Reward and Final Answer Confidence Reward-we show that R-TAP-enhanced models consistently outperform conventional single-pass methods for both large language models (LLMs) and vision-language models (VLMs). Moreover, by analyzing the frequency of "Oops"-like expressions in model responses, we find that R-TAP-applied models exhibit significantly fewer self-reflective patterns, resulting in more stable and faster inference-time reasoning. We hope R-TAP pave the way evolving into efficient and elaborated methods to refine the reasoning processes of future AI.