Finding Answers in Thought Matters: Revisiting Evaluation on Large Language Models with Reasoning

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Jo, Hwiyeol, Lee, Joosung, Lee, Jaehone, Lee, Sang-Woo, Park, Joonsuk, Yoo, Kang Min
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915558630359040
author Jo, Hwiyeol
Lee, Joosung
Lee, Jaehone
Lee, Sang-Woo
Park, Joonsuk
Yoo, Kang Min
author_facet Jo, Hwiyeol
Lee, Joosung
Lee, Jaehone
Lee, Sang-Woo
Park, Joonsuk
Yoo, Kang Min
contents Evaluating generative models, such as large language models (LLMs), commonly involves question-answering tasks where the final answer is selected based on probability of answer choices. On the other hand, for models requiring reasoning, the method of answer extraction plays a critical role. Our research reveals that the performance of reasoning models and their final answer distributions are highly sensitive to the answer extraction algorithm employed. In order to mitigate this, we propose a basic framework: Answer Regeneration. The method uses an additional model inference, providing the prior input and output prefaced by the prompt "Answer:". The final answer is then selected or extracted from the regenerated output. We show that this extraction-rule-agnostic approach exhibits improved performance and enhanced robustness. Furthermore, we have applied this framework to general math problems and open-ended question answering tasks. Our analysis and this framework could offer a more reliable results for model evaluation.
format Preprint
id arxiv_https___arxiv_org_abs_2510_14773
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Finding Answers in Thought Matters: Revisiting Evaluation on Large Language Models with Reasoning
Jo, Hwiyeol
Lee, Joosung
Lee, Jaehone
Lee, Sang-Woo
Park, Joonsuk
Yoo, Kang Min
Computation and Language
Artificial Intelligence
Evaluating generative models, such as large language models (LLMs), commonly involves question-answering tasks where the final answer is selected based on probability of answer choices. On the other hand, for models requiring reasoning, the method of answer extraction plays a critical role. Our research reveals that the performance of reasoning models and their final answer distributions are highly sensitive to the answer extraction algorithm employed. In order to mitigate this, we propose a basic framework: Answer Regeneration. The method uses an additional model inference, providing the prior input and output prefaced by the prompt "Answer:". The final answer is then selected or extracted from the regenerated output. We show that this extraction-rule-agnostic approach exhibits improved performance and enhanced robustness. Furthermore, we have applied this framework to general math problems and open-ended question answering tasks. Our analysis and this framework could offer a more reliable results for model evaluation.
title Finding Answers in Thought Matters: Revisiting Evaluation on Large Language Models with Reasoning
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2510.14773