VORTEXRAG: 7-Layer RAG โ€” Causal Drift Filtering + Context Poison Guard [paper + code + demo]

#186
by vigneshwar234 - opened

Relevant for anyone using this model as a retrieval backbone.

One limitation of all embedding-based retrieval is that cosine similarity can't separate causal relevance from topical association. VORTEXRAG addresses this by adding causal filtering on top of embedding retrieval.

Architecture: your embedding model handles ANN search โ†’ VORTEXRAG's SDC/CPG layers filter by causal drift โ†’ FV layer verifies faithfulness post-generation.

Results: EM 74.8, Faithfulness 0.94 (+0.23 over standard embedding retrieval baseline). 11 domain presets (medical ฯ„=0.35, legal ฯ„=0.40, scientific ฯ„=0.30) for plug-and-play deployment.

Paper: https://doi.org/10.5281/zenodo.20579702
Code (MIT, 229 tests): https://github.com/vignesh2027/VORTEXRAG
Demo: https://huggingface.co/spaces/vigneshwar234/VORTEXRAG

KennethEnevoldsen changed discussion status to closed

Sign up or log in to comment