Thailand

As the scaling of Large Language Models (LLMs) has dramatically enhanced their capabilities, there has been a growing focus on the alignment problem to ensure their responsible and ethical use. While existing alignment efforts predominantly concentrate on universal values such as the HHH principle, the aspect of culture, which is inherently pluralistic and diverse, has not received adequate attention. This work introduces a new benchmark, CDEval, aimed at evaluating the cultural dimensions of LLMs. CDEval is constructed by incorporating both GPT-4&#39;s automated generation and human verification, covering six cultural dimensions across seven domains. Our comprehensive experiments provide intriguing insights into the culture of mainstream LLMs, highlighting both consistencies and variations across different dimensions and domains. The findings underscore the importance of integrating cultural considerations in LLM development, particularly for applications in diverse cultural settings. This benchmark serves as a valuable resource for cultural studies in LLMs, paving the way for more culturally aware and sensitive models.

ACL 2024

CDEval: A Benchmark for Measuring the Cultural Dimensions of Large Language Models

large language models；cultural dimension；benchmark； cultural analysis

As the scaling of Large Language Models (LLMs) has dramatically enhanced their capabilities, there has been a growing focus on the alignment problem to ensure their responsible and ethical use. While existing alignment efforts predominantly concentrate on universal values such as the HHH principle, the aspect of culture, which is inherently pluralistic and diverse, has not received adequate attention. This work introduces a new benchmark, CDEval, aimed at evaluating the cultural dimensions of LLMs. CDEval is constructed by incorporating both GPT-4's automated generation and human verification, covering six cultural dimensions across seven domains. Our comprehensive experiments provide intriguing insights into the culture of mainstream LLMs, highlighting both consistencies and variations across different dimensions and domains. The findings underscore the importance of integrating cultural considerations in LLM development, particularly for applications in diverse cultural settings. This benchmark serves as a valuable resource for cultural studies in LLMs, paving the way for more culturally aware and sensitive models.

workshop paper

### Welcome!
The 62nd Annual Meeting of the Association for Computational Linguistics (ACL 2024) will take place in Bangkok, Thailand from August 11th to 16th, 2024. Our Virtual Poster Sessions will take place online Thursday, August 22, 2024.

You are required to register for this event. **Please register [here](https://2024.aclweb.org/registration). **

If you have already registered, please check your inbox for an email from Underline granting you access to ACL 2024 content.

Please register!

The 62nd Annual Meeting of the Association for Computational Linguistics (ACL 2024) will take place in Bangkok, Thailand from August 11th to 16th, 2024. More information will be announced soon.

This study explores the sources of instability in maintaining cultural personas and opinions within multi-agent LLM systems. Drawing on simulations of inter-cultural collaboration and debate, we analyze agents' pre- and post-discussion private responses alongside chat transcripts to assess the stability of cultural personas and the impact of opinion diversity on group outcomes. Our findings suggest that multi-agent discussions can encourage collective decisions that reflect diverse perspectives, yet this benefit is tempered by the agents' susceptibility to conformity due to perceived peer pressure and challenges in maintaining consistent personas and opinions. Counterintuitively, instructions that encourage debate in support of one's opinions increase the rate of instability. Without addressing the factors we identify, the full potential of multi-agent frameworks for producing more culturally diverse AI outputs will remain untapped.

Conformity, Confabulation, and Impersonation: Persona Inconstancy in Multi-Agent LLM Collaboration

The difference in culture between the U.S. and Japan is a popular subject for Western vs. Eastern cultural comparison for researchers. One particular challenge is to obtain and annotate multilingual datasets. In this study, we utilized COVID-19 tweets from the two countries as a case study, focusing particularly on discussions concerning masks. The annotation task was designed to gain insights into societal attitudes toward the mask policies implemented in both countries. The aim of this study is to provide a practical approach for the annotation task by thoroughly documenting how we aligned the multilingual annotation guidelines to obtain a comparable dataset. We proceeded to document the effective practices during our annotation process to synchronize our multilingual guidelines. Furthermore, we discussed difficulties caused by differences in expression style and culture, and potential strategies that helped improve our agreement scores and reduce discrepancies between the annotation results in both languages. These findings offer an alternative method for synchronizing multilingual annotation guidelines and achieving feasible agreement scores for cross-cultural annotation tasks. This study resulted in a multilingual guideline in English and Japanese to annotate topics related to public discourses about COVID-19 masks in the U.S. and Japan.

Synchronizing Approach in Designing Annotation Guidelines for Multilingual Datasets: A COVID-19 Case Study Using English and Japanese Tweets

Large language models (LLMs) have rapidly evolved as the foundation of various natural language processing (NLP) applications. Despite their wide use cases, their understanding of culturally-related concepts and reasoning remains limited. Meantime, there is a significant need to enhance these models' cultural reasoning capabilities, especially concerning underrepresented regions. This paper introduces a novel pipeline for extracting high-quality, culturally-related instruction tuning datasets from vast unstructured corpora. We utilize a self-instruction generation pipeline to identify cultural concepts and trigger instruction. By integrating with a general-purpose instruction tuning dataset, our model demonstrates enhanced capabilities in recognizing and understanding regional cultural nuances, thereby enhancing its reasoning capabilities. We conduct experiments across three regions: Singapore, the Philippines, and the United States, achieving performance improvement of up to 6%. Our research opens new avenues for extracting cultural instruction tuning sets directly from unstructured data, setting a precedent for future innovations in the field.

CRAFT: Extracting and Tuning Cultural Instructions from the Wild

Alignment of the language model with human preferences is a common approach to making a language model useful to end users.
However, most alignment work is done in English, and human preference datasets are dominated by English, reflecting only the preferences of English-speaking annotators.
Nevertheless, it is common practice to use the English preference data, either directly or by translating it into the target language, when aligning a multilingual language model.
The question is whether such an alignment strategy marginalizes the preference of non-English speaking users.
To this end, we investigate the effect of aligning Japanese language models with (mostly) English resources.
In particular, we focus on evaluating whether the commonsense morality of the resulting fine-tuned models is aligned with Japanese culture using the JCommonsenseMorality (JCM) and ETHICS datasets.
The experimental results show that the fine-tuned model outperforms the SFT model. However, it does not demonstrate the same level of improvement as a model fine-tuned using the JCM, suggesting that while some aspects of commonsense morality are transferable, others may not be.

Does Cross-Cultural Alignment Change the Commonsense Morality of Language Models?

Integrating Large Language Models (LLMs) into various global cultures presents a cultural challenge: they must respect social norms, and avoid transgressing cultural boundaries. This calls for LLMs to be able to $\textit{adapt}$ their outputs to diverse cultural norms. However, the extent of this $\textit{cultural adaptability}$ remains unclear. To this end, we introduce $NormAd$, a dataset of 2.6k stories that represent social and cultural norms from 75 countries to assess the ability of LLMs to adapt to different granular contexts. Evaluation on $NormAd$ suggests that LLMs struggle with cultural reasoning across all contextual granularities -- Mistral-7b-Instruct, one of the top performing models, achieves only 81.8\% accuracy versus 95.6\% achieved by humans. Additionally, LLMs show stronger adaptability to English-centric cultures over those from the Global South, and find it considerably easier to assess the social acceptability of stories that adhere to cultural norms than those that deviate from them. Our benchmark emphasizes the potential to make models more equitable towards global audiences.

$NormAd$: A Benchmark for Measuring the Cultural Adaptability of Large Language Models

While preliminary findings indicate that multilingual LLMs exhibit reduced bias compared to monolingual ones, a comprehensive understanding of the effect of multilingual training on bias mitigation, is lacking. This study addresses this gap by systematically training six LLMs of identical size (2.6B parameters) and architecture: five monolingual models (English, German, French, Italian, and Spanish) and one multilingual model trained on an equal distribution of data across these languages, all using publicly available data. To ensure robust evaluation, standard bias benchmarks were automatically translated into the five target languages and verified for both translation quality and bias preservation by human annotators. Our results consistently demonstrate that multilingual training effectively mitigates bias. Moreover, we observe that multilingual models achieve not only lower bias but also superior prediction accuracy when compared to monolingual models with the same amount of training data, model architecture, and size.

Do Multilingual Large Language Models Mitigate Stereotype Bias?

Large language models (LLMs) often face knowledge conflicts when addressing questions that can yield different answers based on the cultural or regional context. In this work, we create the **QARV** benchmark to evaluate LLM's proficiency in managing these conflicts between two distinct countries, the US and Korea. The benchmark consists of 671 handwritten questions, each accompanied by two answers from each nation. Our evaluation of 42 different LLMs, each tested with nine unique prompts, reveals that most models inherently respond with an American bias. Additionally, we observe that prompting, and instruction-tuning, are all practical measures to enhance cultural alignment.

Assessing LLMs' Ability to Navigate Cultural Knowledge Conflicts

Cultural differences in common ground may result in pragmatic failure and misunderstandings. We develop our method Rational Speech Acts for Cross-Cultural Communication (RSA+C3) to resolve cross-cultural differences in common ground. To measure the success of our method, we study RSA+C3 in the collaborative referential game of Codenames Duet and show that our method successfully improves collaboration between simulated players of different cultures. Our contributions are threefold: (1) creating Codenames players using contrastive learning of an embedding space and LLM prompting that are aligned with human patterns of play, (2) studying culturally induced differences in common ground reflected in our trained models, and (3) demonstrating that our method RSA+C3 can ease cross-cultural communication in gameplay by inferring sociocultural context from interaction.

RSA+C3: Cross-Cultural Communication using RSA in Codenames

Large language models (LLMs) show limitations in comprehending everyday cultural commonsense knowledge in diverse regions, especially in non-English languages, as these are not necessarily written explicitly online. Existing works, however, mainly focus on a single language or rely heavily on online data sources like Wikipedia, which are often included in the training data. To address this, we present BLEnD, a multilingual socio-cultural commonsense benchmark with 13K question samples from carefully hand-crafted templates. The benchmark covers cultural aspects across 9 countries and regions (Azerbaijan, China, Greece, Indonesia, Mexico, Spain, UK, US, and West Java) in 7 languages (Azerbaijani, Chinese, English, Greek, Indonesian, Spanish, and Sundanese). We show that LLMs perform better in high-resource cultures, with a maximum 27.9%p difference in GPT-4 under a short-answer question format. Furthermore, they perform better for mid-to-high-resource cultures in their local languages, while for low-resource cultures, they perform better in English.

BLEnD: A Benchmark for LLMs on Everyday Knowledge in Diverse Cultures and Languages

Recent studies have highlighted the presence of cultural biases in Large Language Models (LLMs), yet often lack a robust methodology to dissect these phenomena comprehensively. Our work aims to bridge this gap by delving into the Food domain—a universally relevant yet culturally diverse aspect of human life. We introduce FmLAMA, a multilingual dataset centered on food-related cultural facts and variations in food practices. We analyze LLMs across various architectures and configurations, evaluating their performance in both monolingual and multilingual settings. By leveraging templates in six different languages, we investigate how LLMs interact with language-specific and cultural knowledge. Our findings reveal that (1) LLMs demonstrate a pronounced bias towards food knowledge prevalent in the United States; (2) Incorporating relevant cultural context significantly improves LLMs' ability to access cultural knowledge; (3) The efficacy of LLMs in capturing cultural nuances is highly dependent on the interplay between the probing language, the specific model architecture, and the cultural context in question. This research underscores the complexity of integrating cultural understanding into LLMs and emphasizes the importance of culturally diverse datasets to mitigate biases and enhance model performance across different cultural domains.

Premium content

CDEval: A Benchmark for Measuring the Cultural Dimensions of Large Language Models

Downloads

Next from ACL 2024

Conformity, Confabulation, and Impersonation: Persona Inconstancy in Multi-Agent LLM Collaboration

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES