Matrix Representations in Information Retrieval // Omar Khattab // MLOps Podcast #194 clip

MLOps podcast #194 with Omar Khattab, PhD Candidate at Stanford, DSPy: Compiling Declarative Language Model Calls into Self-Improving Pipelines brought to us by @WeightsBiases.

Omar introduces novel ideas in the realm of information retrieval, urging consideration for late interaction strategies and a shift towards refining pipelines rather than fixating on intricate models.

// Abstract
The ML community is rapidly exploring techniques for prompting language models (LMs) and for stacking them into pipelines that solve complex tasks. Unfortunately, existing LM pipelines …

clip dspy ideas information language language model matrix mlops novel phd pipelines podcast retrieval shift stanford strategies