30 October 2025 12:15 - 12:45
Interactive panel | Human oversight in LLMs: Building for safety, alignment, and trust
Aligning large language models to specific objectives demands fine-tuning pipelines that are both precise and scalable.
In this session, panelists will unpack the technical architecture behind these pipelines - covering data selection, model adjustment techniques, and ongoing alignment monitoring - while tackling challenges like preserving generalization and minimizing bias or drift.