Contents Menu Expand Light mode Dark mode Auto light/dark, in light mode Auto light/dark, in dark mode Skip to content
NetForge
Logo
NetForge

Start

  • Overview
  • Getting Started
  • Python
  • Evaluation splits
  • Probes
  • Your agent
  • Capability tests

Guides

  • Scenarios
  • Baselines
  • Offline RL datasets
  • Python vs JAX
  • Diagnostics
  • Difficulty and splits
  • Reproducibility
  • Belief-state graphs
  • Curriculum
  • Ray RLlib
  • Custom scenarios
  • Verification
  • PettingZoo and Gymnasium
  • SOC export (OCSF)

Benchmarks

  • Benchmarks
  • Baselines
  • Self-play / Elo
  • Run

Architecture

  • Tick loop
  • Dynamic topologies
  • OT/SCADA physics
  • Multi-modal obs
  • Zero-Trust Identity
  • Sim2Real Bridge
  • NLP-SIEM

Cyber

  • Cyber
  • Threat model
  • Red actions
  • Blue actions

API

  • Python API
  • Environment API
  • SIEM API
  • NLP encoder API
  • Sim2Real API

Contribute

  • Contribute
  • Directions

Project

  • Roadmap
  • Scope
  • Datasheet
  • Changelog
Back to top
View this page
Edit this page
Home Start Python Cyber Eval Architecture API GitHub

Benchmarks¶

benchmarks/ scripts, benchmarks/results/ artifacts.

  • Baselines

  • Self-play / Elo

  • Run — throughput, eval, sweeps, env spec

JAX CPU, batch 4096: 270,795 env-steps/s (1,083,181 agent-steps/s). Reproduce in Run.

Next
Baselines
Previous
SOC export (OCSF)
Copyright © 2026, Igor Jankowski
Made with Furo