2.3k citations · 2.4k across the 19 of their papers we have counts for
21 papers
Mecha-nudges for Machines
Giulio Frey, Kawin Ethayarajh
AI agents are becoming active decision-makers on the Internet. As they make decisions in the same environments as humans, the environments themselves can change to influence them.…
Humanline: Online Alignment as Perceptual Loss
Sijia Liu, Niklas Muennighoff, Kawin Ethayarajh
Online alignment (e.g., GRPO) is generally more performant than offline alignment (e.g., DPO) -- but why? Drawing on prospect theory from behavioral economics, we propose a human-c…
Data Checklist: On Unit-Testing Datasets with Usable Information
Heidi C. Zhang, Shabnam Behzad, Kawin Ethayarajh +1
Model checklists (Ribeiro et al., 2020) have emerged as a useful tool for understanding the behavior of LLMs, analogous to unit-testing in software engineering. However, despite da…
KTO: Model Alignment as Prospect Theoretic Optimization
Kawin Ethayarajh, Winnie Xu, Niklas Muennighoff +2
Kahneman & Tversky's tells us that humans perceive random variables in a biased but well-defined manner (1992); for example, humans are famously loss-ave…
Anchor Points: Benchmarking Models with Much Fewer Examples
Rajan Vivek, Kawin Ethayarajh, Diyi Yang +1
Modern language models often exhibit powerful but brittle behavior, leading to the development of larger and more diverse benchmarks to reliably assess their behavior. Here, we sug…
Richer Countries and Richer Representations
Kaitlyn Zhou, Kawin Ethayarajh, Dan Jurafsky
We examine whether some countries are more richly represented in embedding space than others. We find that countries whose names occur with low frequency in training corpora are mo…
Problems with Cosine as a Measure of Embedding Similarity for High Frequency Words
Kaitlyn Zhou, Kawin Ethayarajh, Dallas Card +1
Cosine similarity of contextual embeddings is used in many NLP tasks (e.g., QA, IR, MT) and metrics (e.g., BERTScore). Here, we uncover systematic ways in which word similarities e…
The Authenticity Gap in Human Evaluation
Kawin Ethayarajh, Dan Jurafsky
Human ratings are the gold standard in NLG evaluation. The standard protocol is to collect ratings of generated text, average across annotators, and rank NLG systems by their avera…
Understanding Dataset Difficulty with -Usable Information
Kawin Ethayarajh, Yejin Choi, Swabha Swayamdipta
Estimating the difficulty of a dataset typically involves comparing state-of-the-art models to humans; the bigger the performance gap, the harder the dataset is said to be. However…
Conditional probing: measuring usable information beyond a baseline
John Hewitt, Kawin Ethayarajh, Percy Liang +1
Probing experiments investigate the extent to which neural representations make properties -- like part-of-speech -- predictable. One suggests that a representation encodes a prope…
On the Opportunities and Risks of Foundation Models
Rishi Bommasani, Drew A. Hudson, Ehsan Adeli +111
AI is undergoing a paradigm shift with the rise of models (e.g., BERT, DALL-E, GPT-3) that are trained on broad data at scale and are adaptable to a wide range of downstream tasks.…
Attention Flows are Shapley Value Explanations
Kawin Ethayarajh, Dan Jurafsky
Shapley Values, a solution to the credit assignment problem in cooperative game theory, are a popular type of explanation in machine learning, having been used to explain the impor…
Dynaboard: An Evaluation-As-A-Service Platform for Holistic Next-Generation Benchmarking
Zhiyi Ma, Kawin Ethayarajh, Tristan Thrush +6
We introduce Dynaboard, an evaluation-as-a-service framework for hosting benchmarks and conducting holistic model comparison, integrated with the Dynabench platform. Our platform e…
Frequency-based Distortions in Contextualized Word Embeddings
Kaitlyn Zhou, Kawin Ethayarajh, Dan Jurafsky
How does word frequency in pre-training data affect the behavior of similarity metrics in contextualized BERT embeddings? Are there systematic ways in which some word relationships…
Utility is in the Eye of the User: A Critique of NLP Leaderboards
Kawin Ethayarajh, Dan Jurafsky
Benchmarks such as GLUE have helped drive advances in NLP by incentivizing the creation of more accurate models. While this leaderboard paradigm has been remarkably successful, a h…
Is Your Classifier Actually Biased? Measuring Fairness under Uncertainty with Bernstein Bounds
Kawin Ethayarajh
Most NLP datasets are not annotated with protected attributes such as gender, making it difficult to measure classification bias using standard measures of fairness (e.g., equal op…
BLEU Neighbors: A Reference-less Approach to Automatic Evaluation
Kawin Ethayarajh, Dorsa Sadigh
Evaluation is a bottleneck in the development of natural language generation (NLG) models. Automatic metrics such as BLEU rely on references, but for tasks such as open-ended gener…
Rotate King to get Queen: Word Relationships as Orthogonal Transformations in Embedding Space
Kawin Ethayarajh
A notable property of word embeddings is that word relationships can exist as linear substructures in the embedding space. For example, corresponds to $\vec{\text…
How Contextual are Contextualized Word Representations? Comparing the Geometry of BERT, ELMo, and GPT-2 Embeddings
Kawin Ethayarajh
Replacing static word embeddings with contextualized word representations has yielded significant improvements on many NLP tasks. However, just how contextual are the contextualize…
Understanding Undesirable Word Embedding Associations
Kawin Ethayarajh, David Duvenaud, Graeme Hirst
Word embeddings are often criticized for capturing undesirable word associations such as gender stereotypes. However, methods for measuring and removing such biases remain poorly u…
Towards Understanding Linear Word Analogies
Kawin Ethayarajh, David Duvenaud, Graeme Hirst
A surprising property of word vectors is that word analogies can often be solved with vector arithmetic. However, it is unclear why arithmetic operators correspond to non-linear em…