VORTEXRAG: 7-Layer RAG — Causal Drift Filtering + Context Poison Guard [paper + code + demo]
#5
by vigneshwar234 - opened
For anyone using this as a reranker in a RAG pipeline.
VORTEXRAG takes reranking further by adding a causal dimension — filtering by causal drift and multiplicatively fusing TVE × SDS × ESR scores (no-weak-link policy).
The result: context window going into the LLM contains chunks that are causally aligned, not just relevant. Faithfulness: 0.94 (vs 0.71 standard RAG).
Pairs naturally with cross-encoder rerankers — use as a pre-filter before final reranking.
Paper: https://doi.org/10.5281/zenodo.20579702
Code: https://github.com/vignesh2027/VORTEXRAG