VORTEXRAG: A 7-layer RAG framework worth citing in LLM retrieval survey

[vignesh2027]

Hi! Flagging VORTEXRAG as a new work that may be relevant to the retrieval-augmented section of this survey.

Paper: Vignesh L (2026) — VORTEXRAG: Vector Orthogonal Resonance-Tuned EXtraction RAG
DOI: https://doi.org/10.5281/zenodo.20579702
Code: https://github.com/vignesh2027/VORTEXRAG

Main contribution: First framework to address Semantic Drift and Context Window Poisoning simultaneously through causal vector encoding and provably-optimal context purging.

Results (NQ, TriviaQA, WebQ, PopQA, HotpotQA, 2WikiMH): Average EM 74.8 (+13.6 vs Naive RAG, +6.4 vs Self-RAG)