Sharon Levy
Research Subjects
Sharon Levy has a total of 32 co-authors. Here are the top co-authors by number of shared publications:
- William Yang Wang
University of California Santa Barbara
11 shared publications - Alex Mei
University of California Santa Barbara Santa Barbara CA
4 shared publications - Aesha Parekh
University of California Santa Barbara
2 shared publications - Diba Mirza
University of California Santa Barbara Santa Barbara CA
2 shared publications - Lily Ou
University of California Santa Barbara
2 shared publications - Michael Saxon
University of California Santa Barbara
2 shared publications
Latest Publications
- Open Access
ASSERT: Automated Safety Scenario Red Teaming for Evaluating the Robustness of Large Language Models
As large language models are integrated into society, robustness toward a suite of prompts is increasingly important to maintain reliability in a high-variance environment.Robustness evaluations must comprehensively encapsulate the various settings in which a user may invoke an intelligent system. This paper proposes ASSERT, Automated Safety Scenario Red Teaming, consisting of three methods -- semantically aligned augmentation, target bootstrapping, and adversarial knowledge injection. For robust safety evaluation, we apply these methods in the critical domain of AI safety to algorithmically generate a test suite of prompts covering diverse robustness settings -- semantic equivalence, related scenarios, and adversarial. We partition our prompts into four safety domains for a fine-grained analysis of how the domain affects model performance. Despite dedicated safeguards in existing state-of-the-art models, we find statistically significant performance differences of up to 11% in absolute classification accuracy among semantically related scenarios and error rates of up to 19% absolute error in zero-shot adversarial settings, raising concerns for users' physical safety.
Open AccessASSERT: Automated Safety Scenario Red Teaming for Evaluating the Robustness of Large Language Models
As large language models are integrated into society, robustness toward a suite of prompts is increasingly important to maintain reliability in a high-variance environment.Robustness evaluations must comprehensively encapsulate the various settings in which a user may invoke an intelligent system. This paper proposes ASSERT, Automated Safety Scenario Red Teaming, consisting of three methods -- semantically aligned augmentation, target bootstrapping, and adversarial knowledge injection. For robust safety evaluation, we apply these methods in the critical domain of AI safety to algorithmically generate a test suite of prompts covering diverse robustness settings -- semantic equivalence, related scenarios, and adversarial. We partition our prompts into four safety domains for a fine-grained analysis of how the domain affects model performance. Despite dedicated safeguards in existing state-of-the-art models, we find statistically significant performance differences of up to 11% in absolute classification accuracy among semantically related scenarios and error rates of up to 19% absolute error in zero-shot adversarial settings, raising concerns for users' physical safety.
- Open Access
Foveate, Attribute, and Rationalize: Towards Physically Safe and Trustworthy AI
Users' physical safety is an increasing concern as the market for intelligent systems continues to grow, where unconstrained systems may recommend users dangerous actions that can lead to serious injury. Covertly unsafe text is an area of particular interest, as such text may arise from everyday scenarios and are challenging to detect as harmful. We propose FARM, a novel framework leveraging external knowledge for trustworthy rationale generation in the context of safety. In particular, FARM foveates on missing knowledge to qualify the information required to reason in specific scenarios and retrieves this information with attribution to trustworthy sources. This knowledge is used to both classify the safety of the original text and generate human-interpretable rationales, shedding light on the risk of systems to specific user groups and helping both stakeholders manage the risks of their systems and policymakers to provide concrete safeguards for consumer safety. Our experiments show that FARM obtains state-of-the-art results on the SafeText dataset, showing absolute improvement in safety classification accuracy by 5.9%.
Open AccessFoveate, Attribute, and Rationalize: Towards Physically Safe and Trustworthy AI
Users' physical safety is an increasing concern as the market for intelligent systems continues to grow, where unconstrained systems may recommend users dangerous actions that can lead to serious injury. Covertly unsafe text is an area of particular interest, as such text may arise from everyday scenarios and are challenging to detect as harmful. We propose FARM, a novel framework leveraging external knowledge for trustworthy rationale generation in the context of safety. In particular, FARM foveates on missing knowledge to qualify the information required to reason in specific scenarios and retrieves this information with attribution to trustworthy sources. This knowledge is used to both classify the safety of the original text and generate human-interpretable rationales, shedding light on the risk of systems to specific user groups and helping both stakeholders manage the risks of their systems and policymakers to provide concrete safeguards for consumer safety. Our experiments show that FARM obtains state-of-the-art results on the SafeText dataset, showing absolute improvement in safety classification accuracy by 5.9%.
- Open Access
Comparing Biases and the Impact of Multilingual Training across Multiple Languages
Studies in bias and fairness in natural language processing have primarily examined social biases within a single language and/or across few attributes (e.g. gender, race). However, biases can manifest differently across various languages for individual attributes. As a result, it is critical to examine biases within each language and attribute. Of equal importance is to study how these biases compare across languages and how the biases are affected when training a model on multilingual data versus monolingual data. We present a bias analysis across Italian, Chinese, English, Hebrew, and Spanish on the downstream sentiment analysis task to observe whether specific demographics are viewed more positively. We study bias similarities and differences across these languages and investigate the impact of multilingual vs. monolingual training data. We adapt existing sentiment bias templates in English to Italian, Chinese, Hebrew, and Spanish for four attributes: race, religion, nationality, and gender. Our results reveal similarities in bias expression such as favoritism of groups that are dominant in each language's culture (e.g. majority religions and nationalities). Additionally, we find an increased variation in predictions across protected groups, indicating bias amplification, after multilingual finetuning in comparison to multilingual pretraining.
Open AccessComparing Biases and the Impact of Multilingual Training across Multiple Languages
Studies in bias and fairness in natural language processing have primarily examined social biases within a single language and/or across few attributes (e.g. gender, race). However, biases can manifest differently across various languages for individual attributes. As a result, it is critical to examine biases within each language and attribute. Of equal importance is to study how these biases compare across languages and how the biases are affected when training a model on multilingual data versus monolingual data. We present a bias analysis across Italian, Chinese, English, Hebrew, and Spanish on the downstream sentiment analysis task to observe whether specific demographics are viewed more positively. We study bias similarities and differences across these languages and investigate the impact of multilingual vs. monolingual training data. We adapt existing sentiment bias templates in English to Italian, Chinese, Hebrew, and Spanish for four attributes: race, religion, nationality, and gender. Our results reveal similarities in bias expression such as favoritism of groups that are dominant in each language's culture (e.g. majority religions and nationalities). Additionally, we find an increased variation in predictions across protected groups, indicating bias amplification, after multilingual finetuning in comparison to multilingual pretraining.
- Open Access
Mitigating Covertly Unsafe Text within Natural Language Systems
An increasingly prevalent problem for intelligent technologies is text safety, as uncontrolled systems may generate recommendations to their users that lead to injury or life-threatening consequences. However, the degree of explicitness of a generated statement that can cause physical harm varies. In this paper, we distinguish types of text that can lead to physical harm and establish one particularly underexplored category: covertly unsafe text. Then, we further break down this category with respect to the system's information and discuss solutions to mitigate the generation of text in each of these subcategories. Ultimately, our work defines the problem of covertly unsafe language that causes physical harm and argues that this subtle yet dangerous issue needs to be prioritized by stakeholders and regulators. We highlight mitigation strategies to inspire future researchers to tackle this challenging problem and help improve safety within smart systems.
Open AccessMitigating Covertly Unsafe Text within Natural Language Systems
An increasingly prevalent problem for intelligent technologies is text safety, as uncontrolled systems may generate recommendations to their users that lead to injury or life-threatening consequences. However, the degree of explicitness of a generated statement that can cause physical harm varies. In this paper, we distinguish types of text that can lead to physical harm and establish one particularly underexplored category: covertly unsafe text. Then, we further break down this category with respect to the system's information and discuss solutions to mitigate the generation of text in each of these subcategories. Ultimately, our work defines the problem of covertly unsafe language that causes physical harm and argues that this subtle yet dangerous issue needs to be prioritized by stakeholders and regulators. We highlight mitigation strategies to inspire future researchers to tackle this challenging problem and help improve safety within smart systems.
- Open Access
- 1 January 2023
- Published by Association for Computational Linguistics (ACL)
- p. 5831-5847
ASSERT: Automated Safety Scenario Red Teaming for Evaluating the Robustness of Large Language Models
Alex Mei, Sharon Levy, William Wang. Findings of the Association for Computational Linguistics: EMNLP 2023. 2023.
- 1 January 2023
- Published by Association for Computational Linguistics (ACL)
- p. 5831-5847
Open AccessASSERT: Automated Safety Scenario Red Teaming for Evaluating the Robustness of Large Language Models
Alex Mei, Sharon Levy, William Wang. Findings of the Association for Computational Linguistics: EMNLP 2023. 2023.
- Open Access
- 1 January 2023
- Published by Association for Computational Linguistics (ACL)
- p. 1-10
Addressing Issues of Cross-Linguality in Open-Retrieval Question Answering Systems For Emergent Domains
Alon Albalak, Sharon Levy, William Yang Wang. Proceedings of the 17th Conference of the European Chapter of the Association for Computational Linguistics: System Demonstrations. 2023.
- 1 January 2023
- Published by Association for Computational Linguistics (ACL)
- p. 1-10
Open AccessAddressing Issues of Cross-Linguality in Open-Retrieval Question Answering Systems For Emergent Domains
Alon Albalak, Sharon Levy, William Yang Wang. Proceedings of the 17th Conference of the European Chapter of the Association for Computational Linguistics: System Demonstrations. 2023.
- Open Access
Modeling Disclosive Transparency in NLP Application Descriptions
Broader disclosive transparency$-$truth and clarity in communication regarding the function of AI systems$-$is widely considered desirable. Unfortunately, it is a nebulous concept, difficult to both define and quantify. This is problematic, as previous work has demonstrated possible trade-offs and negative consequences to disclosive transparency, such as a confusion effect, where "too much information" clouds a reader's understanding of what a system description means. Disclosive transparency's subjective nature has rendered deep study into these problems and their remedies difficult. To improve this state of affairs, We introduce neural language model-based probabilistic metrics to directly model disclosive transparency, and demonstrate that they correlate with user and expert opinions of system transparency, making them a valid objective proxy. Finally, we demonstrate the use of these metrics in a pilot study quantifying the relationships between transparency, confusion, and user perceptions in a corpus of real NLP system descriptions.
Open AccessModeling Disclosive Transparency in NLP Application Descriptions
Broader disclosive transparency$-$truth and clarity in communication regarding the function of AI systems$-$is widely considered desirable. Unfortunately, it is a nebulous concept, difficult to both define and quantify. This is problematic, as previous work has demonstrated possible trade-offs and negative consequences to disclosive transparency, such as a confusion effect, where "too much information" clouds a reader's understanding of what a system description means. Disclosive transparency's subjective nature has rendered deep study into these problems and their remedies difficult. To improve this state of affairs, We introduce neural language model-based probabilistic metrics to directly model disclosive transparency, and demonstrate that they correlate with user and expert opinions of system transparency, making them a valid objective proxy. Finally, we demonstrate the use of these metrics in a pilot study quantifying the relationships between transparency, confusion, and user perceptions in a corpus of real NLP system descriptions.
- Open Access
Investigating Memorization of Conspiracy Theories in Text Generation
The adoption of natural language generation (NLG) models can leave individuals vulnerable to the generation of harmful information memorized by the models, such as conspiracy theories. While previous studies examine conspiracy theories in the context of social media, they have not evaluated their presence in the new space of generative language models. In this work, we investigate the capability of language models to generate conspiracy theory text. Specifically, we aim to answer: can we test pretrained generative language models for the memorization and elicitation of conspiracy theories without access to the model's training data? We highlight the difficulties of this task and discuss it in the context of memorization, generalization, and hallucination. Utilizing a new dataset consisting of conspiracy theory topics and machine-generated conspiracy theories helps us discover that many conspiracy theories are deeply rooted in the pretrained language models. Our experiments demonstrate a relationship between model parameters such as size and temperature and their propensity to generate conspiracy theory text. These results indicate the need for a more thorough review of NLG applications before release and an in-depth discussion of the drawbacks of memorization in generative language models.
Open AccessInvestigating Memorization of Conspiracy Theories in Text Generation
The adoption of natural language generation (NLG) models can leave individuals vulnerable to the generation of harmful information memorized by the models, such as conspiracy theories. While previous studies examine conspiracy theories in the context of social media, they have not evaluated their presence in the new space of generative language models. In this work, we investigate the capability of language models to generate conspiracy theory text. Specifically, we aim to answer: can we test pretrained generative language models for the memorization and elicitation of conspiracy theories without access to the model's training data? We highlight the difficulties of this task and discuss it in the context of memorization, generalization, and hallucination. Utilizing a new dataset consisting of conspiracy theory topics and machine-generated conspiracy theories helps us discover that many conspiracy theories are deeply rooted in the pretrained language models. Our experiments demonstrate a relationship between model parameters such as size and temperature and their propensity to generate conspiracy theory text. These results indicate the need for a more thorough review of NLG applications before release and an in-depth discussion of the drawbacks of memorization in generative language models.
- Open Access
Investigating African-American Vernacular English in Transformer-Based Text Generation
The growth of social media has encouraged the written use of African American Vernacular English (AAVE), which has traditionally been used only in oral contexts. However, NLP models have historically been developed using dominant English varieties, such as Standard American English (SAE), due to text corpora availability. We investigate the performance of GPT-2 on AAVE text by creating a dataset of intent-equivalent parallel AAVE/SAE tweet pairs, thereby isolating syntactic structure and AAVE- or SAE-specific language for each pair. We evaluate each sample and its GPT-2 generated text with pretrained sentiment classifiers and find that while AAVE text results in more classifications of negative sentiment than SAE, the use of GPT-2 generally increases occurrences of positive sentiment for both. Additionally, we conduct human evaluation of AAVE and SAE text generated with GPT-2 to compare contextual rigor and overall quality.
Open AccessInvestigating African-American Vernacular English in Transformer-Based Text Generation
The growth of social media has encouraged the written use of African American Vernacular English (AAVE), which has traditionally been used only in oral contexts. However, NLP models have historically been developed using dominant English varieties, such as Standard American English (SAE), due to text corpora availability. We investigate the performance of GPT-2 on AAVE text by creating a dataset of intent-equivalent parallel AAVE/SAE tweet pairs, thereby isolating syntactic structure and AAVE- or SAE-specific language for each pair. We evaluate each sample and its GPT-2 generated text with pretrained sentiment classifiers and find that while AAVE text results in more classifications of negative sentiment than SAE, the use of GPT-2 generally increases occurrences of positive sentiment for both. Additionally, we conduct human evaluation of AAVE and SAE text generated with GPT-2 to compare contextual rigor and overall quality.
- Open Access
Cross-lingual Transfer Learning for COVID-19 Outbreak Alignment
The spread of COVID-19 has become a significant and troubling aspect of society in 2020. With millions of cases reported across countries, new outbreaks have occurred and followed patterns of previously affected areas. Many disease detection models do not incorporate the wealth of social media data that can be utilized for modeling and predicting its spread. In this case, it is useful to ask, can we utilize this knowledge in one country to model the outbreak in another? To answer this, we propose the task of cross-lingual transfer learning for epidemiological alignment. Utilizing both macro and micro text features, we train on Italy's early COVID-19 outbreak through Twitter and transfer to several other countries. Our experiments show strong results with up to 0.85 Spearman correlation in cross-country predictions.
Open AccessCross-lingual Transfer Learning for COVID-19 Outbreak Alignment
The spread of COVID-19 has become a significant and troubling aspect of society in 2020. With millions of cases reported across countries, new outbreaks have occurred and followed patterns of previously affected areas. Many disease detection models do not incorporate the wealth of social media data that can be utilized for modeling and predicting its spread. In this case, it is useful to ask, can we utilize this knowledge in one country to model the outbreak in another? To answer this, we propose the task of cross-lingual transfer learning for epidemiological alignment. Utilizing both macro and micro text features, we train on Italy's early COVID-19 outbreak through Twitter and transfer to several other countries. Our experiments show strong results with up to 0.85 Spearman correlation in cross-country predictions.