LMaaS-Papers

Why we included this project

Anyone working with large language models they can only reach through an API will find this list useful. It collects papers on adapting LLMs to downstream tasks without access to model weights or gradients, which is the constraint you actually live with when the model is a hosted service. The papers are grouped by the techniques that matter in that setting: text prompting, in-context learning, black-box optimization, feature-based learning, and using the model to generate training data. Maintainer Tianxiang Sun keeps a clear scope and notes where the topic shades into prompt-based learning, so the curation has a point of view instead of just piling up NLP papers. With around 67 entries, it is a good starting bibliography for surveying how to get value out of a hosted model.