White papers & research
Technical insights into binary neural networks and efficient AI architectures
Hybrid Binary Intelligence: Neural Networks, Constraints, and Multi-Paradigm Architectures
Binary neural networks replace floating-point arithmetic with single-bit operations, enabling neural network inference at a fraction of conventional computational cost. However, pure binary networks struggle to match the accuracy of full-precision models. This paper presents a hybrid architecture that combines binary neural networks with constraint-based reasoning, hyperdimensional computing, and adaptive precision to achieve greater than 99% accuracy on domain-specific tasks.
By Marc Filipan
Permuted Agreement Popcount: Structural Similarity in Binary Vector Spaces
Hamming distance provides efficient binary vector comparison but fails to capture structural coherence in matching patterns. Two binary vectors with identical Hamming similarity can exhibit fundamentally different spatial arrangements: one with matches forming coherent blocks, another with matches scattered randomly. This structural blindness creates problems in applications where spatial arrangement matters (computer vision, document similarity, binary neural networks).
By Marc Filipan
AI Safety: Beyond the Theater - Real Solutions to Current Threats
Most AI companies claim GDPR compliance, robust security, and EU AI Act readiness. Yet prompt injection attacks succeed against over 90% of commercial LLM platforms. PII leaks from systems displaying compliance badges. Model extraction proceeds undetected. The gap between AI safety claims and actual implementation has become a credibility crisis.
By Marc Filipan
Join the Dweve Waitlist
Get early access to AI that respects your privacy, the planet, and your wallet.
Early Access Benefits
Quick Process
Takes less than 2 minutes • No credit card required
Stay updated on our research
Get notified when we publish new white papers and research findings