Alert button
Picture for Eric Michael Smith

Eric Michael Smith

Alert button

Meta AI

ROBBIE: Robust Bias Evaluation of Large Generative Language Models

Nov 29, 2023
David Esiobu, Xiaoqing Tan, Saghar Hosseini, Megan Ung, Yuchen Zhang, Jude Fernandes, Jane Dwivedi-Yu, Eleonora Presani, Adina Williams, Eric Michael Smith

As generative large language models (LLMs) grow more performant and prevalent, we must develop comprehensive enough tools to measure and improve their fairness. Different prompt-based datasets can be used to measure social bias across multiple text domains and demographic axes, meaning that testing LLMs on more datasets can potentially help us characterize their biases more fully, and better ensure equal and equitable treatment of marginalized demographic groups. In this work, our focus is two-fold: (1) Benchmarking: a comparison of 6 different prompt-based bias and toxicity metrics across 12 demographic axes and 5 families of generative LLMs. Out of those 6 metrics, AdvPromptSet and HolisticBiasR are novel datasets proposed in the paper. The comparison of those benchmarks gives us insights about the bias and toxicity of the compared models. Therefore, we explore the frequency of demographic terms in common LLM pre-training corpora and how this may relate to model biases. (2) Mitigation: we conduct a comprehensive study of how well 3 bias/toxicity mitigation techniques perform across our suite of measurements. ROBBIE aims to provide insights for practitioners while deploying a model, emphasizing the need to not only measure potential harms, but also understand how they arise by characterizing the data, mitigate harms once found, and balance any trade-offs. We open-source our analysis code in hopes of encouraging broader measurements of bias in future LLMs.

* EMNLP 2023 
Viaarxiv icon

The Gender-GAP Pipeline: A Gender-Aware Polyglot Pipeline for Gender Characterisation in 55 Languages

Aug 31, 2023
Benjamin Muller, Belen Alastruey, Prangthip Hansanti, Elahe Kalbassi, Christophe Ropers, Eric Michael Smith, Adina Williams, Luke Zettlemoyer, Pierre Andrews, Marta R. Costa-jussà

Figure 1 for The Gender-GAP Pipeline: A Gender-Aware Polyglot Pipeline for Gender Characterisation in 55 Languages
Figure 2 for The Gender-GAP Pipeline: A Gender-Aware Polyglot Pipeline for Gender Characterisation in 55 Languages
Figure 3 for The Gender-GAP Pipeline: A Gender-Aware Polyglot Pipeline for Gender Characterisation in 55 Languages
Figure 4 for The Gender-GAP Pipeline: A Gender-Aware Polyglot Pipeline for Gender Characterisation in 55 Languages

Gender biases in language generation systems are challenging to mitigate. One possible source for these biases is gender representation disparities in the training and evaluation data. Despite recent progress in documenting this problem and many attempts at mitigating it, we still lack shared methodology and tooling to report gender representation in large datasets. Such quantitative reporting will enable further mitigation, e.g., via data augmentation. This paper describes the Gender-GAP Pipeline (for Gender-Aware Polyglot Pipeline), an automatic pipeline to characterize gender representation in large-scale datasets for 55 languages. The pipeline uses a multilingual lexicon of gendered person-nouns to quantify the gender representation in text. We showcase it to report gender representation in WMT training data and development data for the News task, confirming that current data is skewed towards masculine representation. Having unbalanced datasets may indirectly optimize our systems towards outperforming one gender over the others. We suggest introducing our gender quantification pipeline in current datasets and, ideally, modifying them toward a balanced representation.

* 15 pages 
Viaarxiv icon

Llama 2: Open Foundation and Fine-Tuned Chat Models

Jul 19, 2023
Hugo Touvron, Louis Martin, Kevin Stone, Peter Albert, Amjad Almahairi, Yasmine Babaei, Nikolay Bashlykov, Soumya Batra, Prajjwal Bhargava, Shruti Bhosale, Dan Bikel, Lukas Blecher, Cristian Canton Ferrer, Moya Chen, Guillem Cucurull, David Esiobu, Jude Fernandes, Jeremy Fu, Wenyin Fu, Brian Fuller, Cynthia Gao, Vedanuj Goswami, Naman Goyal, Anthony Hartshorn, Saghar Hosseini, Rui Hou, Hakan Inan, Marcin Kardas, Viktor Kerkez, Madian Khabsa, Isabel Kloumann, Artem Korenev, Punit Singh Koura, Marie-Anne Lachaux, Thibaut Lavril, Jenya Lee, Diana Liskovich, Yinghai Lu, Yuning Mao, Xavier Martinet, Todor Mihaylov, Pushkar Mishra, Igor Molybog, Yixin Nie, Andrew Poulton, Jeremy Reizenstein, Rashi Rungta, Kalyan Saladi, Alan Schelten, Ruan Silva, Eric Michael Smith, Ranjan Subramanian, Xiaoqing Ellen Tan, Binh Tang, Ross Taylor, Adina Williams, Jian Xiang Kuan, Puxin Xu, Zheng Yan, Iliyan Zarov, Yuchen Zhang, Angela Fan, Melanie Kambadur, Sharan Narang, Aurelien Rodriguez, Robert Stojnic, Sergey Edunov, Thomas Scialom

Figure 1 for Llama 2: Open Foundation and Fine-Tuned Chat Models
Figure 2 for Llama 2: Open Foundation and Fine-Tuned Chat Models
Figure 3 for Llama 2: Open Foundation and Fine-Tuned Chat Models
Figure 4 for Llama 2: Open Foundation and Fine-Tuned Chat Models

In this work, we develop and release Llama 2, a collection of pretrained and fine-tuned large language models (LLMs) ranging in scale from 7 billion to 70 billion parameters. Our fine-tuned LLMs, called Llama 2-Chat, are optimized for dialogue use cases. Our models outperform open-source chat models on most benchmarks we tested, and based on our human evaluations for helpfulness and safety, may be a suitable substitute for closed-source models. We provide a detailed description of our approach to fine-tuning and safety improvements of Llama 2-Chat in order to enable the community to build on our work and contribute to the responsible development of LLMs.

Viaarxiv icon

Improving Open Language Models by Learning from Organic Interactions

Jun 07, 2023
Jing Xu, Da Ju, Joshua Lane, Mojtaba Komeili, Eric Michael Smith, Megan Ung, Morteza Behrooz, William Ngan, Rashel Moritz, Sainbayar Sukhbaatar, Y-Lan Boureau, Jason Weston, Kurt Shuster

Figure 1 for Improving Open Language Models by Learning from Organic Interactions
Figure 2 for Improving Open Language Models by Learning from Organic Interactions
Figure 3 for Improving Open Language Models by Learning from Organic Interactions
Figure 4 for Improving Open Language Models by Learning from Organic Interactions

We present BlenderBot 3x, an update on the conversational model BlenderBot 3, which is now trained using organic conversation and feedback data from participating users of the system in order to improve both its skills and safety. We are publicly releasing the participating de-identified interaction data for use by the research community, in order to spur further progress. Training models with organic data is challenging because interactions with people "in the wild" include both high quality conversations and feedback, as well as adversarial and toxic behavior. We study techniques that enable learning from helpful teachers while avoiding learning from people who are trying to trick the model into unhelpful or toxic responses. BlenderBot 3x is both preferred in conversation to BlenderBot 3, and is shown to produce safer responses in challenging situations. While our current models are still far from perfect, we believe further improvement can be achieved by continued use of the techniques explored in this work.

Viaarxiv icon

BlenderBot 3: a deployed conversational agent that continually learns to responsibly engage

Aug 10, 2022
Kurt Shuster, Jing Xu, Mojtaba Komeili, Da Ju, Eric Michael Smith, Stephen Roller, Megan Ung, Moya Chen, Kushal Arora, Joshua Lane, Morteza Behrooz, William Ngan, Spencer Poff, Naman Goyal, Arthur Szlam, Y-Lan Boureau, Melanie Kambadur, Jason Weston

Figure 1 for BlenderBot 3: a deployed conversational agent that continually learns to responsibly engage
Figure 2 for BlenderBot 3: a deployed conversational agent that continually learns to responsibly engage
Figure 3 for BlenderBot 3: a deployed conversational agent that continually learns to responsibly engage
Figure 4 for BlenderBot 3: a deployed conversational agent that continually learns to responsibly engage

We present BlenderBot 3, a 175B parameter dialogue model capable of open-domain conversation with access to the internet and a long-term memory, and having been trained on a large number of user defined tasks. We release both the model weights and code, and have also deployed the model on a public web page to interact with organic users. This technical report describes how the model was built (architecture, model and training scheme), and details of its deployment, including safety mechanisms. Human evaluations show its superiority to existing open-domain dialogue agents, including its predecessors (Roller et al., 2021; Komeili et al., 2022). Finally, we detail our plan for continual learning using the data collected from deployment, which will also be publicly released. The goal of this research program is thus to enable the community to study ever-improving responsible agents that learn through interaction.

Viaarxiv icon

"I'm sorry to hear that": finding bias in language models with a holistic descriptor dataset

May 18, 2022
Eric Michael Smith, Melissa Hall Melanie Kambadur, Eleonora Presani, Adina Williams

Figure 1 for "I'm sorry to hear that": finding bias in language models with a holistic descriptor dataset
Figure 2 for "I'm sorry to hear that": finding bias in language models with a holistic descriptor dataset
Figure 3 for "I'm sorry to hear that": finding bias in language models with a holistic descriptor dataset
Figure 4 for "I'm sorry to hear that": finding bias in language models with a holistic descriptor dataset

As language models grow in popularity, their biases across all possible markers of demographic identity should be measured and addressed in order to avoid perpetuating existing societal harms. Many datasets for measuring bias currently exist, but they are restricted in their coverage of demographic axes, and are commonly used with preset bias tests that presuppose which types of biases the models exhibit. In this work, we present a new, more inclusive dataset, HOLISTICBIAS, which consists of nearly 600 descriptor terms across 13 different demographic axes. HOLISTICBIAS was assembled in conversation with experts and community members with lived experience through a participatory process. We use these descriptors combinatorially in a set of bias measurement templates to produce over 450,000 unique sentence prompts, and we use these prompts to explore, identify, and reduce novel forms of bias in several generative models. We demonstrate that our dataset is highly efficacious for measuring previously unmeasurable biases in token likelihoods and generations from language models, as well as in an offensiveness classifier. We will invite additions and amendments to the dataset, and we hope it will help serve as a basis for easy-to-use and more standardized methods for evaluating bias in NLP models.

Viaarxiv icon

Human Evaluation of Conversations is an Open Problem: comparing the sensitivity of various methods for evaluating dialogue agents

Jan 12, 2022
Eric Michael Smith, Orion Hsu, Rebecca Qian, Stephen Roller, Y-Lan Boureau, Jason Weston

Figure 1 for Human Evaluation of Conversations is an Open Problem: comparing the sensitivity of various methods for evaluating dialogue agents
Figure 2 for Human Evaluation of Conversations is an Open Problem: comparing the sensitivity of various methods for evaluating dialogue agents
Figure 3 for Human Evaluation of Conversations is an Open Problem: comparing the sensitivity of various methods for evaluating dialogue agents
Figure 4 for Human Evaluation of Conversations is an Open Problem: comparing the sensitivity of various methods for evaluating dialogue agents

At the heart of improving conversational AI is the open problem of how to evaluate conversations. Issues with automatic metrics are well known (Liu et al., 2016, arXiv:1603.08023), with human evaluations still considered the gold standard. Unfortunately, how to perform human evaluations is also an open problem: differing data collection methods have varying levels of human agreement and statistical sensitivity, resulting in differing amounts of human annotation hours and labor costs. In this work we compare five different crowdworker-based human evaluation methods and find that different methods are best depending on the types of models compared, with no clear winner across the board. While this highlights the open problems in the area, our analysis leads to advice of when to use which one, and possible future directions.

Viaarxiv icon

Hi, my name is Martha: Using names to measure and mitigate bias in generative dialogue models

Sep 07, 2021
Eric Michael Smith, Adina Williams

Figure 1 for Hi, my name is Martha: Using names to measure and mitigate bias in generative dialogue models
Figure 2 for Hi, my name is Martha: Using names to measure and mitigate bias in generative dialogue models
Figure 3 for Hi, my name is Martha: Using names to measure and mitigate bias in generative dialogue models
Figure 4 for Hi, my name is Martha: Using names to measure and mitigate bias in generative dialogue models

All AI models are susceptible to learning biases in data that they are trained on. For generative dialogue models, being trained on real human conversations containing unbalanced gender and race/ethnicity references can lead to models that display learned biases, which we define here broadly as any measurable differences in the distributions of words or semantic content of conversations based on demographic groups. We measure the strength of such biases by producing artificial conversations between two copies of a dialogue model, conditioning one conversational partner to state a name commonly associated with a certain gender and/or race/ethnicity. We find that larger capacity models tend to exhibit more gender bias and greater stereotyping of occupations by gender. We show that several methods of tuning these dialogue models, specifically name scrambling, controlled generation, and unlikelihood training, are effective in reducing bias in conversation, including on a downstream conversational task. Name scrambling is also effective in lowering differences in token usage across conversations where partners have names associated with different genders or races/ethnicities.

Viaarxiv icon

Multi-Modal Open-Domain Dialogue

Oct 02, 2020
Kurt Shuster, Eric Michael Smith, Da Ju, Jason Weston

Figure 1 for Multi-Modal Open-Domain Dialogue
Figure 2 for Multi-Modal Open-Domain Dialogue
Figure 3 for Multi-Modal Open-Domain Dialogue
Figure 4 for Multi-Modal Open-Domain Dialogue

Recent work in open-domain conversational agents has demonstrated that significant improvements in model engagingness and humanness metrics can be achieved via massive scaling in both pre-training data and model size (Adiwardana et al., 2020; Roller et al., 2020). However, if we want to build agents with human-like abilities, we must expand beyond handling just text. A particularly important topic is the ability to see images and communicate about what is perceived. With the goal of engaging humans in multi-modal dialogue, we investigate combining components from state-of-the-art open-domain dialogue agents with those from state-of-the-art vision models. We study incorporating different image fusion schemes and domain-adaptive pre-training and fine-tuning strategies, and show that our best resulting model outperforms strong existing models in multi-modal dialogue while simultaneously performing as well as its predecessor (text-only) BlenderBot (Roller et al., 2020) in text-based conversation. We additionally investigate and incorporate safety components in our final model, and show that such efforts do not diminish model performance with respect to engagingness metrics.

Viaarxiv icon