specific method · filed under evaluation
ProtocolQA robustness validation
Checking protocol-troubleshooting results through refusal checks, additional training data, and an uncontaminated related evaluation.
Also called ProtocolQA robustness checks.
- 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.
We validated our protocol troubleshooting results by checking that the model never refused, adding more protocol-debugging training data, and adding a new protocol-troubleshooting eval similar to ProtocolQA but uncontaminated.
usedevaluation onlyin gpt-oss-120b and gpt-oss-20bOpenAI
Filed alongside
Other methods under evaluation.
Prompt-based output standardizationEvaluation without safety filtersRefusal-suppressed variants for capability estimationavg@kAlmost@1Behavior testing with stubs and edge casesDeadline-bounded rollout evaluationDiverse benchmark validation for quantizationDocumenting evaluation configurationEnd-to-end exploit development evaluationEvaluation equivalence thresholdFixed MTP draft-token acceptance lengthFour-run mean pass@1Full trajectory releaseJoint validation protocolMean@5Model–harness co-designOfficial-release-source baseline restrictionPrompt engineering to elicit answersRefusal behavior quantificationRollout-based auditingSample test and length-constraint filteringSandboxed GPU kernel optimization evaluationSingle-agent and multi-agent evaluation configurations