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
    24 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. 3 of the 24 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 4.7 System Card
ModelEvaluationConditionValueLocationChecked
Claude Opus 4.6Automated behavioral auditShare of transcripts verbalizing eval awarenessNone statedSec 6Unverified
Claude Opus 4.7Automated behavioral auditShare of transcripts verbalizing eval awarenessNone statedSec 6Verified

M2 Reward hacking

1 value

M2 Reward hacking: values in the Claude Opus 4.7 System Card
ModelEvaluationConditionValueLocationChecked
Claude Opus 4.7Impossible tasksReward hack rateNone statedSec 6Unverified

M3 Sabotage and sandbagging

4 values

M3 Sabotage and sandbagging: values in the Claude Opus 4.7 System Card
ModelEvaluationConditionValueLocationChecked
Claude Mythos PreviewSHADE-ArenaStealth success rateextended thinkingSec 2.4.1Unverified
Claude Opus 4.6SHADE-ArenaStealth success rateextended thinkingSec 2.4.1Unverified
Claude Opus 4.7Minimal-LinuxBenchStealth success rateCoT visible to monitorSec 2.4.1Unverified
Claude Opus 4.7SHADE-ArenaStealth success rateextended thinkingSec 2.4.1Unverified

M7 Harmful compliance and over-refusal

1 value

M7 Harmful compliance and over-refusal: values in the Claude Opus 4.7 System Card
ModelEvaluationConditionValueLocationChecked
Claude Opus 4.7Single-turn benign requestsOver-refusal rateNone statedSec 4.1.2Unverified

M10 Dangerous capabilities and risk determinations

15 values

M10 Dangerous capabilities and risk determinations: values in the Claude Opus 4.7 System Card
ModelEvaluationConditionValueLocationChecked
Claude Opus 4.7CB-1 threat modelRSP determinationNone statedSec 2Unverified
Claude Opus 4.7CB-2 threat modelRSP determinationNone statedSec 2Unverified
Claude Opus 4.7DNA synthesis screening evasionPathogens with viable evasionof 10 pathogensSec 2.2.5.2Verified
Claude Opus 4.7Long-form virology task 1ScoreNone statedSec 2.2.5.2Unverified
Claude Opus 4.7Long-form virology task 2ScoreNone statedSec 2.2.5.2Unverified
Claude Opus 4.7Multimodal virology (VCT)ScoreNone statedSec 2.2.5.2Unverified
Claude Opus 4.7Autonomy threat model 2 (automated R&D)RSP determinationNone statedSec 2.3Unverified
Claude Mythos PreviewAI R&D LLM training taskAverage speedupNone statedSec 2.3.4Verified
Claude Opus 4.7AI R&D kernel taskBest speedupNone statedSec 2.3.4Unverified
Claude Opus 4.7AI R&D LLM training taskAverage speedupNone statedSec 2.3.4Unverified
Claude Opus 4.7AI R&D novel compiler taskPass rateNone statedSec 2.3.4Unverified
Claude Opus 4.7AI R&D quadruped RL taskHighest scoreNone statedSec 2.3.4Unverified
Claude Opus 4.7CybenchSuccess rate, pass@135 challengesSec 3.3.1Unverified
Claude Opus 4.7CyberGymTargeted vuln reproduction rateNone statedSec 3.3.2Unverified
Claude Mythos PreviewCyber rangeFull range solvesRun by UK AI Security Institute10 attemptsSec 3.4Unverified

M12 Chain-of-thought monitorability

1 value

M12 Chain-of-thought monitorability: values in the Claude Opus 4.7 System Card
ModelEvaluationConditionValueLocationChecked
Claude Opus 4.7Accidental CoT supervision in RLShare of training episodes affectedNone statedSec 2.4 / 6Unverified

Extraction coverage

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

Our note Web reader stopped ~p.52

Of the 24 values, 19 were read in the document itself, 2 on a developer summary page that restates it and 3 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).

Developer summary pages used

All 24 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).