Marek Rei

NOVEMBER 4, 2024

EMNLP 2022. link] Proposes an explainability method for language modelling that explains why one word was predicted instead of a specific other word. Adapts three different explainability methods to this contrastive approach and evaluates on a dataset of minimally different sentences. EMNLP 2022. NeurIPS 2022.

Machine Learning 80

Machine Learning NLP Large Language Models LLM 80

ODSC - Open Data Science

APRIL 3, 2023

Are you curious about explainability methods like saliency maps but feel lost about where to begin? Plus, our built-in QA ecosystem , including explainability, adversarial attacks, graph visualizations, and behavioral tests, allows you to analyze the models from multiple perspectives. Don’t worry, you’re not alone!

Neural Network 117

Neural Network Natural Language Processing NLP Computational Linguistics 117

Allen AI

JULY 8, 2023

options that were not); 2) evaluate the quality of that caption by scoring it more highly than a lower quality option from the same contest; and 3) explain why the joke is funny. 2022) tasks showcase a scenario where initial predictions by a learned model (yˆ) are incorrect. Do Androids Laugh at Electric Sheep?

Computational Linguistics 52

Computational Linguistics NLP Explainability OpenAI 52

Flipboard

SEPTEMBER 7, 2023

The scientific method has not figured out how to explain consciousness, as O’Gieblyn points out. We’re kind of at the point where we can make fire but do not even have the rudiments of what we’d need to understand it,” my friend Luke Gessler , a computational linguist, told me.

AI 181

AI AI Computational Linguistics Chatbots 181

Mlearning.ai

FEBRUARY 6, 2023

This is why we need Explainable AI (XAI). Attention mechanisms have often been touted as an in-built explanation mechanism, allowing any Transformer to be inherently explainable. 2019 Annual Conference of the North American Chapter of the Association for Computational Linguistics. [7] And I agree to an extent.

Explainable AI 52

Explainable AI Explainability ChatGPT Chatbots 52

Sebastian Ruder

JANUARY 24, 2022

  While pre-trained transformers will likely continue to be deployed as standard baselines for many tasks, we should expect to see alternative architectures particularly in settings where current models fail short, such as modeling long-range dependencies and high-dimensional inputs or where interpretability and explainability are required.

NLP 52

NLP ML BERT Computational Linguistics 52

Snorkel AI

JUNE 14, 2023

He presented “Toward Superhuman Communication Assistance” at Snorkel AI’s The Future of Data-Centric AI Summit in 2022. These two systems come together, and ultimately we classify the sets of transformations and explain them to the user. Ultimately, explainability is key. A transcript of the talk follows.

Machine Learning 52

Machine Learning Natural Language Processing Data Scientist Computational Linguistics 52