A

adversarial

Veridion System is a modular intelligence infrastructure layer designed to transform fragmented, multi-source data into trusted and explainable signals. It continuously ingests, normalizes, and correlates information while applying integrity verification, adversarial noise detection, and counter-signal validation to ensure robustness. The system produces scored, auditable outputs that can be used by downstream AI systems, analytics platforms, and decision engines operating in uncertain or adversarial information environments. https://roxanneardary.com/veridion-system/

Empirical validation of C4 geometric defense against 16 Agents of Chaos. 550 adversarial prompts. 4 defense systems. 96.7% block rate. LLM validation on GPT-4o-mini + Mistral 7B. MIT.

A minimal Python3 program to collect and store JSON responses concurrently from LLM models using zai-sdk.

The repository contains an Unreal Engine project for a strategy role playing videogame. The game's combat complexity is low-medium as of the current implementation. The project is meant to be used to test search based adversarial algorithms.

Implemented: Alpha-beta, Negascout, MTD(f), MonteCarloTreeSearch(UCT), MCTS(Epsilon-greedy) and RAVE. There is an incomplete version of UCT-RAVE.

An implementation of a Monte Carlo tree search adversarial agent