Explore practical strategies for collecting high-quality RLHF (Reinforcement Learning from Human Feedback) data in this 12-minute conference talk by Andrew Mauboussin from Surge AI. Learn about the risks associated with low-quality data collection for RLHF and discover effective techniques implemented in Surge's full-stack RLHF data collection product. Gain insights from Mauboussin's extensive experience in ML engineering, including his work in the finance industry at Kensho and election safeguarding at Twitter. Understand how Surge AI has been powering ML teams at leading companies like Anthropic and OpenAI through their innovative approach to human feedback data collection at scale.
Overview
Syllabus
RLHF Data Collection in Practice // Andrew Mauboussin // LLMs in Prod Conference Part 2
Taught by
MLOps.community
Related Courses
-
Reinforcement Learning from Human Feedback (RLHF)
-
Reinforcement Learning from Human Feedback (RLHF) - Advances in Generative AI
-
Reinforcement Learning from Human Feedback (RLHF) Explained
-
CriticGPT: Understanding RLHF and Force Sampling Beam Search Optimization
-
End-to-end Modern Machine Learning in Production - Part 2
-
Evaluating LLM-based Applications - Part 2
