​​​The 2024 Conference on Empirical Methods in Natural Language Processing (EMNLP) has accepted our scientific paper

We’ll present our scientific paper at EMNLP 2024, the leading conference in natural language processing (NLP) and artificial intelligence (AI), from November 12 to 16 in Florida

We’re proud to announce that EMNLP 2024 has accepted our scientific paper, “Self-AMPLIFY: Improving Small Language Models with Self Post Hoc Explanations.” EMNLP 2024 is a leading conference in the area of NLP and AI, and generative AI (GenAI) in particular.

The paper was written by Milan Bhan, PhD student and Domain Specialist at Ekimetrics, Nicolas Chesneau, Head of Innovation at Ekimetrics, Jean-Noël Vittaut, and Marie-Jeanne Lesot, both​ from LIP6 lab at Sorbonne Université. Milan, who will present his paper to the scientists attending EMNLP, explains why he wrote it for his PhD in interpretability and NLP:

“I wrote this article as part of my PhD at LIP6 at Sorbonne Université. The aim is to improve the performance of autoregressive language models by using explainability methods.”

The accepted paper introduces Self-AMPLIFY, our three-step method to improve Small Language Models’ performance by automatically generating rationales from post hoc explanation methods. It targets samples, generates rationales, and builds a final prompt to leverage In-Context Learning.

Our research applied to business gives us a head start in GenAI, and our talents allow us to make significant scientific advances. By breaking down a problem before answering, Self-AMPLIFY enhances GenAI algorithms’ performance in various question-answering tasks, providing more precise answers. Self-AMPLIFY is the first approach to enrich the prompt without human-annotated rationales or the use of auxiliary models.

EMNLP 2024 accepting our paper highlights our expertise in natural language processing, specifically in generative ​AI.

Link to the original paper: Self-AMPLIFY: Improving Small Language Models with Self Post Hoc Explanations