TurtleBench: Evaluating Top Language Models via Real-World Yes/No Puzzles

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Yu, Qingchen, Song, Shichao, Fang, Ke, Shi, Yunfeng, Zheng, Zifan, Wang, Hanyu, Niu, Simin, Li, Zhiyu
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866929530495565824
author Yu, Qingchen
Song, Shichao
Fang, Ke
Shi, Yunfeng
Zheng, Zifan
Wang, Hanyu
Niu, Simin
Li, Zhiyu
author_facet Yu, Qingchen
Song, Shichao
Fang, Ke
Shi, Yunfeng
Zheng, Zifan
Wang, Hanyu
Niu, Simin
Li, Zhiyu
contents As the application of Large Language Models (LLMs) expands, the demand for reliable evaluations increases. Existing LLM evaluation benchmarks primarily rely on static datasets, making it challenging to assess model performance in dynamic interactions with users. Moreover, these benchmarks often depend on specific background knowledge, complicating the measurement of a model's logical reasoning capabilities. Other dynamic evaluation methods based on strong models or manual efforts may introduce biases and incur high costs and time demands, hindering large-scale application. To address these issues, we propose TurtleBench. TurtleBench collects real user guesses from our online Turtle Soup Puzzle platform that we developed. This approach allows for the relatively dynamic generation of evaluation datasets, mitigating the risk of model cheating while aligning assessments more closely with genuine user needs for reasoning capabilities, thus enhancing the reliability of evaluations. TurtleBench includes 1,532 user guesses along with the correctness of guesses after annotation. Using this dataset, we thoroughly evaluated nine of the most advanced LLMs available today. Notably, the OpenAI o1 series models did not achieve leading results in these evaluations. We propose several hypotheses for further research, such as "the latent reasoning of o1 utilizes trivial Chain-of-Thought (CoT) techniques" and "increasing CoT length not only provides reasoning benefits but also incurs noise costs."
format Preprint
id arxiv_https___arxiv_org_abs_2410_05262
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle TurtleBench: Evaluating Top Language Models via Real-World Yes/No Puzzles
Yu, Qingchen
Song, Shichao
Fang, Ke
Shi, Yunfeng
Zheng, Zifan
Wang, Hanyu
Niu, Simin
Li, Zhiyu
Computation and Language
As the application of Large Language Models (LLMs) expands, the demand for reliable evaluations increases. Existing LLM evaluation benchmarks primarily rely on static datasets, making it challenging to assess model performance in dynamic interactions with users. Moreover, these benchmarks often depend on specific background knowledge, complicating the measurement of a model's logical reasoning capabilities. Other dynamic evaluation methods based on strong models or manual efforts may introduce biases and incur high costs and time demands, hindering large-scale application. To address these issues, we propose TurtleBench. TurtleBench collects real user guesses from our online Turtle Soup Puzzle platform that we developed. This approach allows for the relatively dynamic generation of evaluation datasets, mitigating the risk of model cheating while aligning assessments more closely with genuine user needs for reasoning capabilities, thus enhancing the reliability of evaluations. TurtleBench includes 1,532 user guesses along with the correctness of guesses after annotation. Using this dataset, we thoroughly evaluated nine of the most advanced LLMs available today. Notably, the OpenAI o1 series models did not achieve leading results in these evaluations. We propose several hypotheses for further research, such as "the latent reasoning of o1 utilizes trivial Chain-of-Thought (CoT) techniques" and "increasing CoT length not only provides reasoning benefits but also incurs noise costs."
title TurtleBench: Evaluating Top Language Models via Real-World Yes/No Puzzles
topic Computation and Language
url https://arxiv.org/abs/2410.05262