Saved in:
| Main Authors: | , , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2024
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2404.05221 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866913464013815808 |
|---|---|
| author | Hao, Shibo Gu, Yi Luo, Haotian Liu, Tianyang Shao, Xiyan Wang, Xinyuan Xie, Shuhua Ma, Haodi Samavedhi, Adithya Gao, Qiyue Wang, Zhen Hu, Zhiting |
| author_facet | Hao, Shibo Gu, Yi Luo, Haotian Liu, Tianyang Shao, Xiyan Wang, Xinyuan Xie, Shuhua Ma, Haodi Samavedhi, Adithya Gao, Qiyue Wang, Zhen Hu, Zhiting |
| contents | Generating accurate step-by-step reasoning is essential for Large Language Models (LLMs) to address complex problems and enhance robustness and interpretability. Despite the flux of research on developing advanced reasoning approaches, systematically analyzing the diverse LLMs and reasoning strategies in generating reasoning chains remains a significant challenge. The difficulties stem from the lack of two key elements: (1) an automatic method for evaluating the generated reasoning chains on different tasks, and (2) a unified formalism and implementation of the diverse reasoning approaches for systematic comparison. This paper aims to close the gap: (1) We introduce AutoRace for fully automated reasoning chain evaluation. Existing metrics rely on expensive human annotations or pre-defined LLM prompts not adaptable to different tasks. In contrast, AutoRace automatically creates detailed evaluation criteria tailored for each task, and uses GPT-4 for accurate evaluation following the criteria. (2) We develop LLM Reasoners, a library for standardized modular implementation of existing and new reasoning algorithms, under a unified formulation of the search, reward, and world model components. With the new evaluation and library, (3) we conduct extensive study of different reasoning approaches (e.g., CoT, ToT, RAP). The analysis reveals interesting findings about different factors contributing to reasoning, including the reward-guidance, breadth-vs-depth in search, world model, and prompt formats, etc. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2404_05221 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | LLM Reasoners: New Evaluation, Library, and Analysis of Step-by-Step Reasoning with Large Language Models Hao, Shibo Gu, Yi Luo, Haotian Liu, Tianyang Shao, Xiyan Wang, Xinyuan Xie, Shuhua Ma, Haodi Samavedhi, Adithya Gao, Qiyue Wang, Zhen Hu, Zhiting Computation and Language Artificial Intelligence Generating accurate step-by-step reasoning is essential for Large Language Models (LLMs) to address complex problems and enhance robustness and interpretability. Despite the flux of research on developing advanced reasoning approaches, systematically analyzing the diverse LLMs and reasoning strategies in generating reasoning chains remains a significant challenge. The difficulties stem from the lack of two key elements: (1) an automatic method for evaluating the generated reasoning chains on different tasks, and (2) a unified formalism and implementation of the diverse reasoning approaches for systematic comparison. This paper aims to close the gap: (1) We introduce AutoRace for fully automated reasoning chain evaluation. Existing metrics rely on expensive human annotations or pre-defined LLM prompts not adaptable to different tasks. In contrast, AutoRace automatically creates detailed evaluation criteria tailored for each task, and uses GPT-4 for accurate evaluation following the criteria. (2) We develop LLM Reasoners, a library for standardized modular implementation of existing and new reasoning algorithms, under a unified formulation of the search, reward, and world model components. With the new evaluation and library, (3) we conduct extensive study of different reasoning approaches (e.g., CoT, ToT, RAP). The analysis reveals interesting findings about different factors contributing to reasoning, including the reward-guidance, breadth-vs-depth in search, world model, and prompt formats, etc. |
| title | LLM Reasoners: New Evaluation, Library, and Analysis of Step-by-Step Reasoning with Large Language Models |
| topic | Computation and Language Artificial Intelligence |
| url | https://arxiv.org/abs/2404.05221 |