June 16-21, 2024 in Mexico City, Mexico

The 2024 Annual Conference of the North American Chapter of the Association for Computational Linguistics (NAACL) is one of the premier conferences for natural language processing research.

Kitware is excited to share our work on enhancing LLMs to support human decisions in an explainable way. We are developing a new large language model (LLM)-based algorithm as part of DARPA’s In the Moment (ITM) effort, laying the groundwork for a trusted AI system that is aligned with a human decision-maker. The algorithm understands natural language descriptions of attributes that reflect societal values (e.g., “fairness” and “mission focus”), predicts how strongly different courses of action align to such attributes, and provides detailed explanations justifying the predictions. Unlike existing methods, the ITM algorithm is generalizable, reduces the need for extensive training samples, and provides detailed explanations.

To hear more about our work, visit us during our in-person poster session or contact our computer vision team.

Poster Session

Language Models are Alignable Decision-Makers: Dataset and Application to the Medical Triage Domain
Tuesday, June 18 from 3:30-5 PM
Session H in Don Diego
Link to paper

Leave a Reply