In this live webinar, taking place on January 21, 2026 11:00 AM EST | 5:00 PM CET, we’ll explore how to identify critical failure modes in neural networks through systematic error slice detection.
Deep-learning models rarely fail at random. In real-world systems, errors tend to concentrate in specific subsets of data due to issues with training, representation gaps, or ambiguity in problem formulation.
This session focuses on practical and research-backed approaches for systematically discovering where models underperform.
We’ll discuss how teams can move beyond manual error inspection toward structured methods that surface meaningful failure patterns tied to data biases, representation, and modeling assumptions.
Why model errors cluster in specific data slices rather than appearing randomly
How identifying critical failure modes changes how models are evaluated and improved
What systematic slice discovery enables for building more robust and trustworthy DL systems
This webinar is designed for data scientists, ML engineers, and algorithm developers working on real-world deep-learning models, and will be hosted by Yotam Azriel, CEO & Co-Founder, and led by Amit Cohen, Data Scientist, and Tom Koren, Head of AI at Tensorleap.