run-benchmark

Fonctions et usage

Runs SpatialVortex benchmarks (flux position accuracy, ELP accuracy, sacred boost verification, geometric reasoning, humanities final exam, performance benchmarks), compares to SOTA/baselines (GPT-4, Claude 3, BERT), generates markdown reports/tables with scores (e.g., 95% accuracy, 111.1% improvement, latency), and suggests optimizations. Automatically invoked for "run benchmarks", "benchmark SpatialVortex", "compare flux accuracy", "eval geometric reasoning", "check ASI progress", "submit results". Uses Rust benchmark suite (cargo run --bin run_benchmarks), saves JSON outputs, and prepares for GitHub PR or PapersWithCode upload.

Démarrage rapide

Suivez les instructions du dépôt pour votre agent. Consultez les fichiers et dépendances avant l’installation.