Enhancing Mathematical Reasoning in LLMs by Stepwise Correction
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2024
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866916442561052672 |
|---|---|
| author | Wu, Zhenyu Zeng, Qingkai Zhang, Zhihan Tan, Zhaoxuan Shen, Chao Jiang, Meng |
| author_facet | Wu, Zhenyu Zeng, Qingkai Zhang, Zhihan Tan, Zhaoxuan Shen, Chao Jiang, Meng |
| contents | Best-of-N decoding methods instruct large language models (LLMs) to generate multiple solutions, score each using a scoring function, and select the highest scored as the final answer to mathematical reasoning problems. However, this repeated independent process often leads to the same mistakes, making the selected solution still incorrect. We propose a novel prompting method named Stepwise Correction (StepCo) that helps LLMs identify and revise incorrect steps in their generated reasoning paths. It iterates verification and revision phases that employ a process-supervised verifier. The verify-then-revise process not only improves answer correctness but also reduces token consumption with fewer paths needed to generate. With StepCo, a series of LLMs demonstrate exceptional performance. Notably, using GPT-4o as the backend LLM, StepCo achieves an average accuracy of 94.1 across eight datasets, significantly outperforming the state-of-the-art Best-of-N method by +2.4, while reducing token consumption by 77.8%. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2410_12934 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Enhancing Mathematical Reasoning in LLMs by Stepwise Correction Wu, Zhenyu Zeng, Qingkai Zhang, Zhihan Tan, Zhaoxuan Shen, Chao Jiang, Meng Computation and Language Best-of-N decoding methods instruct large language models (LLMs) to generate multiple solutions, score each using a scoring function, and select the highest scored as the final answer to mathematical reasoning problems. However, this repeated independent process often leads to the same mistakes, making the selected solution still incorrect. We propose a novel prompting method named Stepwise Correction (StepCo) that helps LLMs identify and revise incorrect steps in their generated reasoning paths. It iterates verification and revision phases that employ a process-supervised verifier. The verify-then-revise process not only improves answer correctness but also reduces token consumption with fewer paths needed to generate. With StepCo, a series of LLMs demonstrate exceptional performance. Notably, using GPT-4o as the backend LLM, StepCo achieves an average accuracy of 94.1 across eight datasets, significantly outperforming the state-of-the-art Best-of-N method by +2.4, while reducing token consumption by 77.8%. |
| title | Enhancing Mathematical Reasoning in LLMs by Stepwise Correction |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2410.12934 |