Guidance Opportunities in Discrete Diffusion

(1) the main activities of the partner
ServiceNow develops a platform for client organizations to manage and automate large-scale processes across various industries. In 2020, ServiceNow acquired Element AI to strengthen its presence in the Artificial Intelligence (AI) research landscape and the Canadian AI ecosystem. This acquisition
enabled the development of AI-driven products that improve the platform’s capabilities. ServiceNow Research has made significant contributions to the field of foundation models, notably in Natural Language Processing (NLP), and has a strong presence in developing generative models for different data domains.
(2) the challenges the partner aims to solve through this project
System prompts can specify the desired behavior of agentic systems in plain English: “Be polite.”, “Be concise.”, “Ground all statements in facts that can be directly quoted from the provided context.”. Does it work? Well, it depends… Current solutions cannot even provide probability estimates for these
requirements to be satisfied, nor a clear understanding of how each such clause affects the outcome. Special instances of this general challenge come up in multimodal and code contexts: this project has a particular interest in those latter aspects.
(3) the anticipated social or economic benefits of the project for the partner organization(s)
Better counterfactual understanding and/or access to calibrated probabilities could greatly improve our service offering, enabling calculated risks, highlighting caveats when necessary, ultimately building trust with the customers and users. Such interpretability features would also improve our capacity to comply with legislation on protected status and otherwise decrease our litigation risks.

Faculty Supervisor:

Tal Arbel

Student:

Partner:

ServiceNow Canada

Discipline:

Computer science

Sector:

Professional, scientific and technical services

University:

McGill University

Program: