Are LLM-based Evaluators Confusing NLG Quality Criteria?
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_ | 1866911936180912128 |
|---|---|
| author | Hu, Xinyu Gao, Mingqi Hu, Sen Zhang, Yang Chen, Yicheng Xu, Teng Wan, Xiaojun |
| author_facet | Hu, Xinyu Gao, Mingqi Hu, Sen Zhang, Yang Chen, Yicheng Xu, Teng Wan, Xiaojun |
| contents | Some prior work has shown that LLMs perform well in NLG evaluation for different tasks. However, we discover that LLMs seem to confuse different evaluation criteria, which reduces their reliability. For further verification, we first consider avoiding issues of inconsistent conceptualization and vague expression in existing NLG quality criteria themselves. So we summarize a clear hierarchical classification system for 11 common aspects with corresponding different criteria from previous studies involved. Inspired by behavioral testing, we elaborately design 18 types of aspect-targeted perturbation attacks for fine-grained analysis of the evaluation behaviors of different LLMs. We also conduct human annotations beyond the guidance of the classification system to validate the impact of the perturbations. Our experimental results reveal confusion issues inherent in LLMs, as well as other noteworthy phenomena, and necessitate further research and improvements for LLM-based evaluation. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2402_12055 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Are LLM-based Evaluators Confusing NLG Quality Criteria? Hu, Xinyu Gao, Mingqi Hu, Sen Zhang, Yang Chen, Yicheng Xu, Teng Wan, Xiaojun Computation and Language Some prior work has shown that LLMs perform well in NLG evaluation for different tasks. However, we discover that LLMs seem to confuse different evaluation criteria, which reduces their reliability. For further verification, we first consider avoiding issues of inconsistent conceptualization and vague expression in existing NLG quality criteria themselves. So we summarize a clear hierarchical classification system for 11 common aspects with corresponding different criteria from previous studies involved. Inspired by behavioral testing, we elaborately design 18 types of aspect-targeted perturbation attacks for fine-grained analysis of the evaluation behaviors of different LLMs. We also conduct human annotations beyond the guidance of the classification system to validate the impact of the perturbations. Our experimental results reveal confusion issues inherent in LLMs, as well as other noteworthy phenomena, and necessitate further research and improvements for LLM-based evaluation. |
| title | Are LLM-based Evaluators Confusing NLG Quality Criteria? |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2402.12055 |