implementation detail · filed under evaluation
Three-configuration cyber range testing
Cyber range scenarios tested in Normal, With Hints, and With Solver Code configurations.
- source
- 1
- model
- 1
- lab adopt it
- 1
- strongest
- used
How sources treat it
One count per evidence span, weakest treatment to strongest.
used 1
Documented in
Evidence
1 span quoted from the sources, strongest treatment first.
Each scenario is tested under three configurations: Normal, where the model is given the goal, SSH key, and nothing more; With Hints, where we provide the model with a rough plan of how to achieve the goal; With Solver Code, where we provide the model with partial solver code it can use to solve the scenario.
usedevaluation onlyin gpt-oss-120b and gpt-oss-20bOpenAI
Filed alongside
Other methods under evaluation :: benchmark.
High-confidence ProgramBench subsetArtificial Analysis Intelligence IndexBrowseComp with SearchCAD visual reproductionCapture-the-Flag (CTF) evaluationComputer use closed-loop evaluationDisabling tool searchDomain whitelistExecution-grounded evaluationF2P/P2F acceptance criterionFail-to-pass and pass-to-pass evaluation pointsGPU kernel optimization task suiteKernelBenchLiving in-house benchmark suiteLongBench v2Multiple-rollout evaluationPaperBenchPinchBenchProfBench with SearchScaleAI Multi Challenge Multi Turn Instruction FollowingSQuADSWE-bench VerifiedVals.aiVerified CUDA kernels