Key Findings Extraction

via “key-finding-extraction-and-structured-summarization”

AI agent for automated systematic literature reviews.

Unique: Uses a multi-stage LLM pipeline with semantic template matching to identify claim-bearing sentences before extraction, then deduplicates findings via embedding-based clustering, rather than extracting all sentences and filtering post-hoc

vs others: More accurate than single-pass LLM extraction because it pre-filters to claim-bearing sentences and uses clustering to identify redundant findings across papers