Greg Kamradt
AI Educator
Verified Proof of Work Artifacts
2 items catalogedEach artifact below represents an authenticated research publication, production code repository, or technical architectural framework directly authored or co-created by Greg Kamradt. Every entry undergoes editorial source verification.
Five Levels of LLM Document Summarization Architecture
Developed the definitive open-source architectural taxonomy for document summarization, categorizing approaches from Stuffing and Map-Reduce to Refine, LangChain Agentic Chunking, and Hierarchical Clustering.
Map-Reduce architectures can lose connective narrative thread across disparate document chapters without cross-chunk summarization passes.
Needle In A Haystack: Pressure Testing Long Context Retrieval in LLMs
Created the industry-standard Needle In A Haystack benchmark that places distinct target facts across varying depths (0% to 100%) and context lengths (1k to 1M+ tokens) to measure retrieval degradation in frontier foundation models.
Single-needle retrieval does not test complex multi-hop reasoning or distributed cross-document synthesis across long contexts.