- This event has passed.
Tools Fail: Detecting Silent Errors in Faulty Tools – Nov. 25, 2024
Session Description
November 25 2024 @ 12:00 pm - 1:00 pm
On November 25th, CARTE Industry Speaker Seminar Series hosts Jimin Sun from Cohere, moderated by Professor Scott Sanner.
Title: Tools Fail: Detecting Silent Errors in Faulty Tools
Abstract: Tools have become a mainstay of Large Language Models (LLMs), allowing them to retrieve knowledge not in their weights, to perform tasks on the web, and even to control robots. However, most ontologies and surveys of tool-use have assumed the core challenge for LLMs is choosing the tool. Instead, we introduce a framework for tools more broadly which guides us to explore a model’s ability to detect “silent” tool errors, and reflect on how to plan. This more directly aligns with the increasingly popular use of models as tools. We provide an initial approach to failure recovery with promising results both on a controlled calculator setting and embodied agent planning.
Speaker Bio: Jimin Sun is a Machine Learning Engineer at Cohere, working on synthetic data for Large Language Models. She is also a 1st year PhD student at the Language Technologies Institute at Carnegie Mellon University supervised by Yonatan Bisk.
Moderator: Scott Sanner, Professor in Industrial Engineering and Cross-appointed in Computer Science at the University of Toronto, and a faculty affiliate of the Vector Institute.
Location: Myhal Centre for Innovation and Entrepreneurship (55 St. George Street), Room 360
Open to all. No registration necessary.
