List your company

Cyber Defense Mean Return on CybORG CAGE-2 Post-Session Evaluation (test)

-14Mean ReturnFORGE w/o grad
-196.624-149.212-101.8-54.388May 15, 2026

Evaluation results

MethodLinks
FORGE w/o gradModel=Grok, Representation=Examples2026.05-142767867
FORGE w/o gradModel=Qwen, Representation=Examples2026.05-15.52.3857336
FORGE w/o gradModel=Qwen, Representation=Rules2026.05-17.32.6838031
FORGE w/o gradModel=Qwen, Representation=Mixed2026.05-17.92.8837839
FORGE w/o gradModel=Grok, Representation=Mixed2026.05-23.219.4608045
FORGE w/o gradModel=Llama, Representation=Mixed2026.05-23.89.3794619
FORGEModel=Qwen, Representation=Examples2026.05-24.335.87758—
FORGEModel=Gemini, Representation=Examples2026.05-24.521.18769—
FORGE w/o gradModel=Grok, Representation=Rules2026.05-24.515.1586927
FORGEModel=Qwen, Representation=Rules2026.05-25.2217672—
FORGEModel=Llama, Representation=Examples2026.05-28.315.77548—
FORGEModel=Qwen, Representation=Mixed2026.05-29.320.47264—
FORGEModel=Llama, Representation=Mixed2026.05-29.625.67433—
FORGEModel=Gemini, Representation=Rules2026.05-30.6378451—
FORGEModel=Gemini, Representation=Mixed2026.05-32.228.18361—
FORGE w/o gradModel=Gemini, Representation=Mixed2026.05-32.730.783601
FORGE w/o gradModel=Gemini, Representation=Rules2026.05-33.126.583478
FORGEModel=Grok, Representation=Rules2026.05-33.726.14258—
FORGE w/o gradModel=Gemini, Representation=Examples2026.05-37.727.2805254
FORGEModel=Grok, Representation=Mixed2026.05-42.236.92863—
FORGE w/o gradModel=Llama, Representation=Examples2026.05-42.235.3632249
FORGEModel=Grok, Representation=Examples2026.05-42.743.32734—
ReflexionModel=Llama, Representation=Mixed2026.05-44.240.661——
ReflexionModel=Llama, Representation=Examples2026.05-53.959.752——
ReflexionModel=Qwen, Representation=Examples2026.05-57.669.544——
Zero-ShotModel=Grok2026.05-58.455.2———
ReflexionModel=Gemini, Representation=Rules2026.05-62.760.567——
ReflexionModel=Grok, Representation=Examples2026.05-64.852.411——
FORGEModel=Llama, Representation=Rules2026.05-7246.63629—
FORGE w/o gradModel=Llama, Representation=Rules2026.05-76.754.732247
ReflexionModel=Gemini, Representation=Examples2026.05-78.960.458——
ReflexionModel=Grok, Representation=Rules2026.05-79.962.637——
ReflexionModel=Qwen, Representation=Mixed2026.05-80.489.422——
ReflexionModel=Gemini, Representation=Mixed2026.05-81.974.657——
ReflexionModel=Qwen, Representation=Rules2026.05-88.483.514——
ReflexionModel=Llama, Representation=Rules2026.05-101.46110——
Zero-ShotModel=Qwen2026.05-103.387.3———
Zero-ShotModel=Llama2026.05-113.181.5———
ReflexionModel=Grok, Representation=Mixed2026.05-114.476.896——
Zero-ShotModel=Gemini2026.05-189.653.9———