Jean-François Delpech

This work presents a training-free, alignment-free framework for corporate intelligence built on deterministic sparse seed vectors, showing how material corporate events emerge as distinct, interpretable semantic profiles, each traceable to the exact source sentences that produced it, with no domain-specific training and no LLM inference.