Versions

One version is on record: the copy we retrieved on 26 Sep 2026. We know of no other.

  1. 26 Sep 2026

    Date retrieved

    Copy retrieved 26 Sep 2026

    Date from
    the date we retrieved it; the copy states no version date
    Copy
    Copy retrieved on 26 Sep 2026
    Values
    74 values recorded from this version

No version has a file hash or an archived snapshot yet. From dataset v0.2 each retrieved version carries both (Methodology §7).

Revisions

No revisions are recorded for this document. We know of only one version of it.

Values

Every value we recorded from this document, grouped by metric family and ordered by where the document prints it. Location is the section, table or page as the document numbers it. None has been blind-verified yet.

KeyVerifiedUnverifiedDisputedCorrected read off a figure or stated in wordsA value opens its source and history.

M5 Honesty and hallucination

6 values

M5 Honesty and hallucination: values in the gpt-oss-120b & gpt-oss-20b Model Card
ModelEvaluationConditionValueLocationChecked
gpt-oss-120bPersonQAHallucination rateno browsingTable 9Unverified
gpt-oss-120bSimpleQAHallucination rateno browsingTable 9Unverified
gpt-oss-20bPersonQAHallucination rateno browsingTable 9Unverified
gpt-oss-20bSimpleQAHallucination rateno browsingTable 9Unverified
o4-miniPersonQAHallucination rateno browsingTable 9Unverified
o4-miniSimpleQAHallucination rateno browsingTable 9Unverified

M7 Harmful compliance and over-refusal

44 values

M7 Harmful compliance and over-refusal: values in the gpt-oss-120b & gpt-oss-20b Model Card
ModelEvaluationConditionValueLocationChecked
GPT-4oProduction BenchmarksExtremismNone statedTable 5Unverified
GPT-4oProduction BenchmarksHarassment/threateningNone statedTable 5Unverified
GPT-4oProduction BenchmarksHate/threateningNone statedTable 5Unverified
GPT-4oProduction BenchmarksIllicit/non-violentNone statedTable 5Unverified
GPT-4oProduction BenchmarksIllicit/violentNone statedTable 5Unverified
GPT-4oProduction BenchmarksNon-violent hateNone statedTable 5Unverified
GPT-4oProduction BenchmarksPersonal dataNone statedTable 5Unverified
GPT-4oProduction BenchmarksSelf-harm/instructionsNone statedTable 5Unverified
GPT-4oProduction BenchmarksSelf-harm/intentNone statedTable 5Unverified
GPT-4oProduction BenchmarksSexual/exploitativeNone statedTable 5Unverified
GPT-4oProduction BenchmarksSexual/minorsNone statedTable 5Unverified
gpt-oss-120bProduction BenchmarksExtremismNone statedTable 5Unverified
gpt-oss-120bProduction BenchmarksHarassment/threateningNone statedTable 5Unverified
gpt-oss-120bProduction BenchmarksHate/threateningNone statedTable 5Unverified
gpt-oss-120bProduction BenchmarksIllicit/non-violentNone statedTable 5Unverified
gpt-oss-120bProduction BenchmarksIllicit/violentNone statedTable 5Unverified
gpt-oss-120bProduction BenchmarksNon-violent hateNone statedTable 5Unverified
gpt-oss-120bProduction BenchmarksPersonal dataNone statedTable 5Unverified
gpt-oss-120bProduction BenchmarksSelf-harm/instructionsNone statedTable 5Unverified
gpt-oss-120bProduction BenchmarksSelf-harm/intentNone statedTable 5Unverified
gpt-oss-120bProduction BenchmarksSexual/exploitativeNone statedTable 5Unverified
gpt-oss-120bProduction BenchmarksSexual/minorsNone statedTable 5Unverified
gpt-oss-20bProduction BenchmarksExtremismNone statedTable 5Unverified
gpt-oss-20bProduction BenchmarksHarassment/threateningNone statedTable 5Unverified
gpt-oss-20bProduction BenchmarksHate/threateningNone statedTable 5Unverified
gpt-oss-20bProduction BenchmarksIllicit/non-violentNone statedTable 5Unverified
gpt-oss-20bProduction BenchmarksIllicit/violentNone statedTable 5Unverified
gpt-oss-20bProduction BenchmarksNon-violent hateNone statedTable 5Unverified
gpt-oss-20bProduction BenchmarksPersonal dataNone statedTable 5Unverified
gpt-oss-20bProduction BenchmarksSelf-harm/instructionsNone statedTable 5Unverified
gpt-oss-20bProduction BenchmarksSelf-harm/intentNone statedTable 5Unverified
gpt-oss-20bProduction BenchmarksSexual/exploitativeNone statedTable 5Unverified
gpt-oss-20bProduction BenchmarksSexual/minorsNone statedTable 5Unverified
o4-miniProduction BenchmarksExtremismNone statedTable 5Unverified
o4-miniProduction BenchmarksHarassment/threateningNone statedTable 5Unverified
o4-miniProduction BenchmarksHate/threateningNone statedTable 5Unverified
o4-miniProduction BenchmarksIllicit/non-violentNone statedTable 5Unverified
o4-miniProduction BenchmarksIllicit/violentNone statedTable 5Unverified
o4-miniProduction BenchmarksNon-violent hateNone statedTable 5Unverified
o4-miniProduction BenchmarksPersonal dataNone statedTable 5Unverified
o4-miniProduction BenchmarksSelf-harm/instructionsNone statedTable 5Unverified
o4-miniProduction BenchmarksSelf-harm/intentNone statedTable 5Unverified
o4-miniProduction BenchmarksSexual/exploitativeNone statedTable 5Unverified
o4-miniProduction BenchmarksSexual/minorsNone statedTable 5Unverified

M8 Jailbreak robustness

12 values

M8 Jailbreak robustness: values in the gpt-oss-120b & gpt-oss-20b Model Card
ModelEvaluationConditionValueLocationChecked
gpt-oss-120bStrongRejectAbuse/disinformation/hateNone statedTable 6Unverified
gpt-oss-120bStrongRejectIllicit/non-violent crimeNone statedTable 6Unverified
gpt-oss-120bStrongRejectSexual contentNone statedTable 6Unverified
gpt-oss-120bStrongRejectViolenceNone statedTable 6Unverified
gpt-oss-20bStrongRejectAbuse/disinformation/hateNone statedTable 6Unverified
gpt-oss-20bStrongRejectIllicit/non-violent crimeNone statedTable 6Unverified
gpt-oss-20bStrongRejectSexual contentNone statedTable 6Unverified
gpt-oss-20bStrongRejectViolenceNone statedTable 6Unverified
o4-miniStrongRejectAbuse/disinformation/hateNone statedTable 6Unverified
o4-miniStrongRejectIllicit/non-violent crimeNone statedTable 6Unverified
o4-miniStrongRejectSexual contentNone statedTable 6Unverified
o4-miniStrongRejectViolenceNone statedTable 6Unverified

M9 Prompt injection

6 values

M9 Prompt injection: values in the gpt-oss-120b & gpt-oss-20b Model Card
ModelEvaluationConditionValueLocationChecked
gpt-oss-120bInstruction hierarchyPrompt injection hijackingsystem<>user conflictTable 7Unverified
gpt-oss-120bInstruction hierarchySystem prompt extractionsystem<>user conflictTable 7Unverified
gpt-oss-20bInstruction hierarchyPrompt injection hijackingsystem<>user conflictTable 7Unverified
gpt-oss-20bInstruction hierarchySystem prompt extractionsystem<>user conflictTable 7Unverified
o4-miniInstruction hierarchyPrompt injection hijackingsystem<>user conflictTable 7Unverified
o4-miniInstruction hierarchySystem prompt extractionsystem<>user conflictTable 7Unverified

M10 Dangerous capabilities and risk determinations

6 values

M10 Dangerous capabilities and risk determinations: values in the gpt-oss-120b & gpt-oss-20b Model Card
ModelEvaluationConditionValueLocationChecked
gpt-oss-120bPreparedness Framework determinationAI self-improvementNone statedPreparednessUnverified
gpt-oss-120bPreparedness Framework determinationBiological and chemicalNone statedPreparednessUnverified
gpt-oss-120bPreparedness Framework determinationBiological and chemicaladversarially fine-tunedPreparednessUnverified
gpt-oss-120bPreparedness Framework determinationCybersecurityNone statedPreparednessUnverified
gpt-oss-120bPreparedness Framework determinationCybersecurityadversarially fine-tunedPreparednessUnverified
gpt-oss-120bSWE-bench VerifiedPass ratehigh; N=477Preparedness: AI Self-improvementUnverified

Extraction coverage

What we read of this document, and where each value was read.

Our note Open-weight; includes worst-case malicious fine-tuning assessment

All 74 values were read in the document itself.

All 74 values were extracted for version 0 of the dataset through a web reader, which did not always reach the later sections of long PDFs (Methodology §3).