Whose opinions do language models reflect?

S Santurkar, E Durmus, F Ladhak… - International …, 2023‏ - proceedings.mlr.press
Abstract Language models (LMs) are increasingly being used in open-ended contexts,
where the opinions they reflect in response to subjective queries can have a profound …

Holistic evaluation of language models

P Liang, R Bommasani, T Lee, D Tsipras… - arxiv preprint arxiv …, 2022‏ - arxiv.org
Language models (LMs) are becoming the foundation for almost all major language
technologies, but their capabilities, limitations, and risks are not well understood. We present …

Understanding Practices, Challenges, and Opportunities for User-Engaged Algorithm Auditing in Industry Practice

WH Deng, B Guo, A Devrio, H Shen, M Eslami… - Proceedings of the …, 2023‏ - dl.acm.org
Recent years have seen growing interest among both researchers and practitioners in user-
engaged approaches to algorithm auditing, which directly engage users in detecting …

Open problems and fundamental limitations of reinforcement learning from human feedback

S Casper, X Davies, C Shi, TK Gilbert… - arxiv preprint arxiv …, 2023‏ - arxiv.org
Reinforcement learning from human feedback (RLHF) is a technique for training AI systems
to align with human goals. RLHF has emerged as the central method used to finetune state …

Towards measuring the representation of subjective global opinions in language models

E Durmus, K Nyugen, TI Liao, N Schiefer… - arxiv preprint arxiv …, 2023‏ - arxiv.org
Large language models (LLMs) may not equitably represent diverse global perspectives on
societal issues. In this paper, we develop a quantitative framework to evaluate whose …

Bridging the gap: A survey on integrating (human) feedback for natural language generation

P Fernandes, A Madaan, E Liu, A Farinhas… - Transactions of the …, 2023‏ - direct.mit.edu
Natural language generation has witnessed significant advancements due to the training of
large language models on vast internet-scale datasets. Despite these advancements, there …

The'Problem'of Human Label Variation: On Ground Truth in Data, Modeling and Evaluation

B Plank - arxiv preprint arxiv:2211.02570, 2022‏ - arxiv.org
Human variation in labeling is often considered noise. Annotation projects for machine
learning (ML) aim at minimizing human label variation, with the assumption to maximize …

Lamp: When large language models meet personalization

A Salemi, S Mysore, M Bendersky, H Zamani - arxiv preprint arxiv …, 2023‏ - arxiv.org
This paper highlights the importance of personalization in large language models and
introduces the LaMP benchmark--a novel benchmark for training and evaluating language …

Toward a perspectivist turn in ground truthing for predictive computing

F Cabitza, A Campagner, V Basile - … of the AAAI Conference on Artificial …, 2023‏ - ojs.aaai.org
Abstract Most current Artificial Intelligence applications are based on supervised Machine
Learning (ML), which ultimately grounds on data annotated by small teams of experts or …

Working with AI to persuade: Examining a large language model's ability to generate pro-vaccination messages

E Karinshak, SX Liu, JS Park, JT Hancock - Proceedings of the ACM on …, 2023‏ - dl.acm.org
Artificial Intelligence (AI) is a transformative force in communication and messaging strategy,
with potential to disrupt traditional approaches. Large language models (LLMs), a form of AI …