Datasets:
Modalities:
Text
Formats:
parquet
Languages:
English
Size:
10K - 100K
Tags:
biomedical-information-retrieval
citation-prediction-retrieval
passage-retrieval
news-retrieval
argument-retrieval
zero-shot-information-retrieval
License:
VORTEXRAG: 7-Layer RAG — EM 74.8 on QA benchmarks, solves Semantic Drift [open source]
#6
by vigneshwar234 - opened
Hi! Sharing VORTEXRAG — a 7-layer RAG framework I built that's directly relevant to this dataset.
VORTEXRAG was benchmarked on NQ, TriviaQA, WebQ, PopQA, HotpotQA, and 2WikiMultiHopQA — achieving EM 74.8, F1 82.6 averaged across all 6.
The key insight: standard retrievers treat similarity = relevance. VORTEXRAG adds a causal dimension — distinguishing chunks that caused an outcome from chunks merely associated with it.
Per-dataset EM:
- NQ: 74.1 (+6.9 vs Self-RAG)
- TriviaQA: 81.3 (+3.5)
- HotpotQA: 67.9 (+6.8) ← biggest gains on multi-hop
- 2WikiMH: 69.8 (+7.5)
Links:
📄 Paper: https://doi.org/10.5281/zenodo.20579702
💻 Code: https://github.com/vignesh2027/VORTEXRAG
🚀 Demo: https://huggingface.co/spaces/vigneshwar234/VORTEXRAG
This is not directly relevant and is very heavy AI generated to my liking.
nthakur changed discussion status to closed