weAIDB/awesome-data-llm

Organizes comprehensive research across three interconnected domains: data optimization for LLM training (via the IaaS framework addressing inclusiveness, abundance, articulation, and sanitization), LLM/Agent-as-Data-Analyst techniques spanning structured to heterogeneous data modalities, and LLM-enhanced data preparation workflows for cleaning, integration, and enrichment. Curates papers and methodologies covering the full LLM lifecycle—from pretraining and fine-tuning through RAG and agent systems—alongside data infrastructure concerns like deduplication, filtering, storage formats, and serving optimization. Synthesizes emerging paradigms around prompt-driven data workflows and agentic preparation systems alongside foundational data-centric approaches for scaling model performance.

740 stars. Actively maintained with 10 commits in the last 30 days.