Robust-GAP: Achieving Zero-Hallucination Causal Summarization in Hierarchical RAG

Short summary

Robust-GAP is a hierarchical RAG framework that eliminates semantic hallucinations in multi-document log summarization by combining dynamic causal graph extraction, schema-guided validation, and provenance-aware metadata propagation. It claims 100% verifiable citations and 0% hallucination rate, improving on prior approaches like Classic GAP. A Python CLI tool is released on GitHub with zero external dependencies.

• •Hierarchical RAG framework using dynamic causal graphs to prevent LLM hallucinations in log summarization
• •Three-stage pipeline: DLCE for causal extraction, SGAV for schema validation, PAPP for provenance tracking
• •Open-source Python CLI released; claims 100% verifiable citations and 0% hallucination rate vs. alternatives

Generated with AI, which can make mistakes.