hallucination-control

Neurosymbolic framework that catches LLM hallucinations mid-reasoning using causal graphs and GNNs. Converts chain-of-thought outputs into Causal Reasoning Graphs, detects flawed steps with a Graph Attention Network, and auto-corrects via RAG context injection ... before errors propagate.

Evidence-grounded medical RAG system that retrieves FDA and NICE drug guidelines, generates cited answers, and safely refuses unsupported queries to minimize hallucinations.

A relational knowledge base of logical relations between concepts (genus-species, part-whole, cause-effect, opposition) across domain universes, designed as a grounding layer for LLMs: machine-readable definitions, validated relation grammar, and an LLM-driven filling engine.

Explicit control and observability over when an LLM should answer, hedge, or refuse — treating generation as a governed system layer, not a side effect of retrieval.

Trustworthy Agentic RAG prototype for scientific knowledge bases with multi-granularity retrieval, answer auditing, and conservative refusal.

Governed AI/MCP delivery-intelligence workbench with synthetic programme data, evidence tracing, evaluation, telemetry, and visual dashboard.