English Español Français Deutsch
← Back to the list
Foundational figures Methodologists Directly upstream of ai100's methodology

Xia Hu

Whether the academic state of the art in language models can be turned into something a practitioner — not a research lab — can actually deploy and trust.

Hu's "Harnessing the Power of Large Language Models in Practice" survey (2023, senior author) sits in a different corner of the survey literature from the academically dense Zhao-or-Wen-style overviews: it organizes the field around questions a real deployment team has to answer — which model fits which problem class, what the data requirements look like at production scale, what fine-tuning versus prompting actually costs, where the trust gaps appear. The same instinct runs through his D2K Lab work on automated and interpretable ML — methodology designed so that someone outside the research lab can reproduce the result and audit how it was reached. For ai100, this is the literature that maps how brand-mention behavior should be evaluated by someone who has to defend the evaluation to a procurement team.

Worth following when
you need to evaluate LLMs from a practitioner's stance and want the literature framed for deployment rather than for benchmark records.
Topics
practical LLM deployment surveys; automated and interpretable ML; LLM efficiency considerations for production use.
Key works
"Harnessing the Power of Large Language Models in Practice: A Survey" (2023, senior author); D2K Lab automated-ML publications; long body of work on interpretable ML and AutoML at Rice.