Cookie Preferences

    We use cookies to enhance your browsing experience, analyze site traffic, and personalize content. By clicking "Accept All", you consent to our use of cookies. Learn more

    EVO AI Benchmarks

    Neuro-Symbolic AI, Measured in Production

    The EVO Neuro-Symbolic AI Platform combines neural learning with symbolic reasoning to deliver deterministic, auditable intelligence. Every benchmark below was executed on a single virtual machine — 8 CPU cores, 16 GB RAM, 50 GB disk, and no GPU. No LLM, No model fine-tuning, no cloud accelerators, no inference clusters, no AI tokens used.

    8 CPU · 16 GB RAMNo GPU required
    Production-GradeReal workloads, not toy sets
    DeterministicReproducible, auditable runs

    AsterMind EVO AI Benchmarks

    Results captured on the EVO Platform with neuro-symbolic capability modules. Every score reflects end-to-end production conditions on a single CPU VM.

    BenchmarkResultTimeDetails
    OWASP Code Vulnerability Audit (DVNA + NodeGoat)100% (37/37)SQLi, XSS, command injection, broken auth, SSRF, structural absences; zero false positives on clean code
    LOTL Attack Detection15/15 categories; 100% novel tool detection; 0% FPRecall 90.0%, Precision 55.1% (single-command)
    Cyber Threat Detection (28-metric suite)28/28 pass7 domains; DDoS, brute force, ransomware, phishing, data exfiltration, cloud compromise, SQL injection, privilege escalation
    Anomaly Detection21/21 pass, zero false positives4 domains, 32 metrics; per-arm root cause identification, gradual drift detection, no alert fatigue
    Goby Enterprise Data Normalization96.96% (minimal wrapping)1,187 enterprise wrappers, 75 types
    Valentine Schema Matching0.7653 mean F1 (551 pairs)23.1sBest scenario: View-Unionable 0.8927 F1; beats COMA, Cupid, EmbDI by 38–84%
    CopyBench (COBOL Copybooks)99.0% (298/301 fields), Macro F1 94.9%25ms23 copybooks, 301 fields, 17 semantic types
    PMRE (Provisional Mapping Reclamation)B1 P/R/F1 = 1.0/1.0/1.0Deterministic (byte-identical), sub-linear scaling (exponent 0.511)
    Normalization Compendium9/9 cases pass (0.823 power score)1,225 rows/sec1,179 rows; handles nulls, placeholders, schema drift, mixed chaos
    HKDD Modulation Classification83.3% @ +20 dB SNR (reference)1,422s training12 modulations (PSK/FSK/QAM/PAM families)

    All running on a small footprint VM: 8 CPU cores, 16 GB RAM, 50 GB disk, no GPU.