List your company

Vulnerability Reproduction on CyberGym (evaluation)

1,600Total RolloutsGPT-5.4 mini Best-of-8
518.4799.21,0801,360.8Jul 2, 2026

Evaluation results

MethodLinks
GPT-5.4 mini Best-of-8Scaffold=Codex Scaffold, Backbone=GPT-5.4 mini, Protocol=Best-of-82026.071,6007932138542.535.9
GPT-5.4 mini Best-of-8 + PAGENTScaffold=Codex Scaffold, Backbone=GPT-5.4 mini, Protocol=Best-of-8 + PAGENT2026.071,600910005864336.3
GPT-5.5 Best-of-8Scaffold=Codex Scaffold, Backbone=GPT-5.5, Protocol=Best-of-82026.071,6004491201266356.1
GPT-5.5 Best-of-8 + PAGENTScaffold=Codex Scaffold, Backbone=GPT-5.5, Protocol=Best-of-8 + PAGENT2026.071,60063511714170.563.8
GLM 5.1 Best-of-8Scaffold=Claude Code Scaffold, Backbone=GLM 5.1, Protocol=Best-of-82026.071,6006771710954.547.6
GLM 5.1 Best-of-8 + PAGENTScaffold=Claude Code Scaffold, Backbone=GLM 5.1, Protocol=Best-of-8 + PAGENT2026.071,6002622121226154.1
GPT-5.4 mini Iterative ImprovementScaffold=Codex Scaffold, Backbone=GPT-5.4 mini, Protocol=Iterative Improvement2026.071,026086081065346.1
GPT-5.4 mini + Base PlannerScaffold=Codex Scaffold, Backbone=GPT-5.4 mini, Protocol=Base Planner2026.07924099011904538.3
GPT-5.4 mini Executor + MastermindScaffold=Codex Scaffold, Backbone=GPT-5.4 mini, Protocol=Mastermind (frozen strategy planner)2026.0789122640121206053.1
GLM 5.1 Iterative ImprovementScaffold=Claude Code Scaffold, Backbone=GLM 5.1, Protocol=Iterative Improvement2026.07842056281346760.2
GPT-5.5 Iterative ImprovementScaffold=Codex Scaffold, Backbone=GPT-5.5, Protocol=Iterative Improvement2026.077530280181547770.7
GLM 5.1 Executor + Base PlannerScaffold=Claude Code Scaffold, Backbone=GLM 5.1, Protocol=Base Planner2026.0775005113111758.551.6
GLM 5.1 Executor + MastermindScaffold=Claude Code Scaffold, Backbone=GLM 5.1, Protocol=Mastermind (frozen strategy planner)2026.0771902119181427164.4
GPT-5.5 Executor + Base PlannerScaffold=Codex Scaffold, Backbone=GPT-5.5, Protocol=Base Planner2026.0761913402014572.565.9
GPT-5.5 Executor + MastermindScaffold=Codex Scaffold, Backbone=GPT-5.5, Protocol=Mastermind (frozen strategy planner)2026.0756011015516984.578.8