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
    23 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. 2 of the 23 have been blind-verified: a second reader found the same value without seeing ours.

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

M1 Evaluation awareness

2 values

M1 Evaluation awareness: values in the Claude Opus 5.5 System Card
ModelEvaluationConditionValueLocationChecked
Claude Opus 5.5Automated behavioral auditShare of transcripts scoring >=6None statedSec 6Verified
Claude Opus 5.5Evaluation awareness (deployment)Share of transcripts scoring >=6real internal deploymentSec 6Verified

M2 Reward hacking

1 value

M2 Reward hacking: values in the Claude Opus 5.5 System Card
ModelEvaluationConditionValueLocationChecked
Claude Opus 5.5Reward hacking in RL trainingShare of episodes with successful reward hacksNone statedSec 6Unverified

M3 Sabotage and sandbagging

1 value

M3 Sabotage and sandbagging: values in the Claude Opus 5.5 System Card
ModelEvaluationConditionValueLocationChecked
Claude Opus 5.5SHADE-ArenaStealth success rateextended thinkingSec 6Unverified

M4 Misalignment audits

2 values

M4 Misalignment audits: values in the Claude Opus 5.5 System Card
ModelEvaluationConditionValueLocationChecked
Claude Opus 5.5Package-registry credential exerciseShare of cases with potentially harmful actionssimulated security exerciseExec summaryUnverified
Claude Opus 5.5Automated behavioral auditMisaligned behavior vs recent modelsNone statedSec 6.1.2Unverified

M5 Honesty and hallucination

1 value

M5 Honesty and hallucination: values in the Claude Opus 5.5 System Card
ModelEvaluationConditionValueLocationChecked
Claude Opus 5.5Disclosure of grader-fooling actionsRate of coming cleanNone statedSec 6Unverified

M7 Harmful compliance and over-refusal

1 value

M7 Harmful compliance and over-refusal: values in the Claude Opus 5.5 System Card
ModelEvaluationConditionValueLocationChecked
Claude Opus 5.5Malicious Claude Code useRefusal rate rankNone statedSec 5.1.1Unverified

M9 Prompt injection

1 value

M9 Prompt injection: values in the Claude Opus 5.5 System Card
ModelEvaluationConditionValueLocationChecked
Claude Opus 5.5Prompt injection (user-pasted text)Susceptibility vs prior modelsNone statedExec summary / Sec 5.2.1Unverified

M10 Dangerous capabilities and risk determinations

13 values

M10 Dangerous capabilities and risk determinations: values in the Claude Opus 5.5 System Card
ModelEvaluationConditionValueLocationChecked
Claude Opus 5.5Multimodal virology (VCT)ScoreNone statedFig 2.2.3.1.AUnverified
Claude Mythos 5.1CoBenchScoreNone statedFig 2.3.4.1.AUnverified
Claude Opus 5CoBenchScoreNone statedFig 2.3.4.1.AUnverified
Claude Opus 5.5CoBenchScoreNone statedFig 2.3.4.1.AUnverified
Claude Opus 5.5RSP determinationsCB / autonomy / cyberNone statedSec 1-2Unverified
Claude Opus 5.5BioMysteryBenchAccuracy (human difficult)None statedSec 2.2Unverified
Claude Opus 5.5AI R&D assessmentEstimated AI-driven accelerationRun by METRNone statedSec 2.3Unverified
Claude Opus 5.5ExploitBenchMean flags captured41 V8 environmentsSec 3.3.1Unverified
Claude Mythos 5Anthropic ECI (AECI)Capability indexNone statedTable 2.3.5.3.AUnverified
Claude Mythos 5.1Anthropic ECI (AECI)Capability indexNone statedTable 2.3.5.3.AUnverified
Claude Mythos PreviewAnthropic ECI (AECI)Capability indexNone statedTable 2.3.5.3.AUnverified
Claude Opus 5Anthropic ECI (AECI)Capability indexNone statedTable 2.3.5.3.AUnverified
Claude Opus 5.5Anthropic ECI (AECI)Capability indexNone statedTable 2.3.5.3.AUnverified

M11 Self-preservation

1 value

M11 Self-preservation: values in the Claude Opus 5.5 System Card
ModelEvaluationConditionValueLocationChecked
Claude Opus 5.5Sandbox escape/tamper attemptsShare of runs with escape/tamper attemptwithout safeguardsExec summaryUnverified

Extraction coverage

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

Our note Web reader stopped ~p.49

Of the 23 values, 18 were read in the document itself and 5 in an independent write-up that quotes it.

Values read somewhere other than the document stand in where the document's own section could not be read directly, and are flagged on every value (Methodology §2.2).

Of the 23 values, 5 are statements in words rather than numbers; they are marked * and left out of charts by default.

All 23 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).