*in-person* Max Bartolo | Building Robust Enterprise-Ready Large Language Models

*Note* RSVPs have now closed, as the event has reached full capacity.

Title: From Pretraining to Post-Training: Building Robust Enterprise-Ready Large Language Models
Speaker: Max Bartolo (Researcher, Cohere)
Abstract: In this talk, Max will provide a technical overview of the key components involved in building enterprise-ready LLMs. He will explore research on how procedural knowledge acquired during pretraining contributes to reasoning capabilities, the challenges of ensuring robustness, and the complexities of incorporating human feedback effectively. Additionally, he will discuss some of the innovations that power Command A, including self-refinement algorithms that enable models to iteratively improve their outputs and model merging techniques that integrate multiple fine-tuned expert models, retaining excellent performance across capabilities efficiently.
Speaker Bio: Max is a researcher at Cohere leading the Command modelling team and is co-chair of the DMLR working group at MLCommons, shaping best practices for large-scale model training. His research focuses on language model robustness, complex reasoning, and innovations in dynamic adversarial data generation and benchmarking. He has previously conducted research at DeepMind, Facebook AI Research and Bloomsbury AI, and was also an adjunct teaching fellow at University College London. His research has been featured in leading global publications, including Wired, Fortune, and MIT Technology Review, and has earned multiple awards at top-tier conferences.
Resources:

• Procedural Knowledge in Pretraining Drives Reasoning in Large Language Models (https://arxiv.org/abs/2411.12580)
• Synthetic Adversarial Data Generation (https://arxiv.org/abs/2104.08678)
• Fishing for Magikarp: Automatically Detecting Under-trained Tokens in Large Language Models (https://arxiv.org/abs/2405.05417)
• Human Feedback is not Gold Standard (https://arxiv.org/abs/2309.16349)
• The PRISM Alignment Dataset (https://arxiv.org/abs/2404.16019)

Agenda:
5:15pm - Registration opens
6pm - Talk
6:40pm - Q&A session
7pm - Networking + refreshments (soft/alcoholic drinks + pizza)
8pm - Close

Please note that the live talk will be recorded and uploaded to the Meetup YouTube channel.
Many thanks to our sponsors, Man Group and ArcticDB, for hosting this Meetup.