SinclairCoder/Instruction-Tuning-Papers

Curated reading list spanning 20+ foundational and recent papers on instruction-tuning approaches for language models. Covers diverse methodologies including multi-task prompted training (T0, FLAN), retrieval-augmented generation, human feedback alignment, and synthetic instruction generation—enabling systematic study of how natural language instructions improve zero-shot and few-shot generalization across NLP tasks. Organized chronologically from 2021-2022, the collection documents the evolution from foundational work to scaling techniques like instruction fine-tuning on 1,000+ tasks and cross-lingual transfer learning.

766 stars. No commits in the last 6 months.