Assessing Dialect Fairness and Robustness of Large Language Models in Reasoning Tasks
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2024
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866908398151270400 |
|---|---|
| author | Lin, Fangru Mao, Shaoguang La Malfa, Emanuele Hofmann, Valentin de Wynter, Adrian Wang, Xun Chen, Si-Qing Wooldridge, Michael Pierrehumbert, Janet B. Wei, Furu |
| author_facet | Lin, Fangru Mao, Shaoguang La Malfa, Emanuele Hofmann, Valentin de Wynter, Adrian Wang, Xun Chen, Si-Qing Wooldridge, Michael Pierrehumbert, Janet B. Wei, Furu |
| contents | Language is not monolithic. While benchmarks, including those designed for multiple languages, are often used as proxies to evaluate the performance of Large Language Models (LLMs), they tend to overlook the nuances of within-language variation and thus fail to model the experience of speakers of non-standard dialects. Focusing on African American Vernacular English (AAVE), we present the first study aimed at objectively assessing the fairness and robustness of LLMs in handling dialects across canonical reasoning tasks, including algorithm, math, logic, and integrated reasoning. We introduce ReDial (Reasoning with Dialect Queries), a benchmark containing 1.2K+ parallel query pairs in Standardized English and AAVE. We hire AAVE speakers, including experts with computer science backgrounds, to rewrite seven popular benchmarks, such as HumanEval and GSM8K. With ReDial, we evaluate widely used LLMs, including GPT, Claude, Llama, Mistral, and the Phi model families. Our findings reveal that almost all of these widely used models show significant brittleness and unfairness to queries in AAVE. Our work establishes a systematic and objective framework for analyzing LLM bias in dialectal queries. Moreover, it highlights how mainstream LLMs provide unfair service to dialect speakers in reasoning tasks, laying a critical foundation for future research. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2410_11005 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Assessing Dialect Fairness and Robustness of Large Language Models in Reasoning Tasks Lin, Fangru Mao, Shaoguang La Malfa, Emanuele Hofmann, Valentin de Wynter, Adrian Wang, Xun Chen, Si-Qing Wooldridge, Michael Pierrehumbert, Janet B. Wei, Furu Computation and Language Machine Learning Language is not monolithic. While benchmarks, including those designed for multiple languages, are often used as proxies to evaluate the performance of Large Language Models (LLMs), they tend to overlook the nuances of within-language variation and thus fail to model the experience of speakers of non-standard dialects. Focusing on African American Vernacular English (AAVE), we present the first study aimed at objectively assessing the fairness and robustness of LLMs in handling dialects across canonical reasoning tasks, including algorithm, math, logic, and integrated reasoning. We introduce ReDial (Reasoning with Dialect Queries), a benchmark containing 1.2K+ parallel query pairs in Standardized English and AAVE. We hire AAVE speakers, including experts with computer science backgrounds, to rewrite seven popular benchmarks, such as HumanEval and GSM8K. With ReDial, we evaluate widely used LLMs, including GPT, Claude, Llama, Mistral, and the Phi model families. Our findings reveal that almost all of these widely used models show significant brittleness and unfairness to queries in AAVE. Our work establishes a systematic and objective framework for analyzing LLM bias in dialectal queries. Moreover, it highlights how mainstream LLMs provide unfair service to dialect speakers in reasoning tasks, laying a critical foundation for future research. |
| title | Assessing Dialect Fairness and Robustness of Large Language Models in Reasoning Tasks |
| topic | Computation and Language Machine Learning |
| url | https://arxiv.org/abs/2410.11005 |