Verbalized grader awareness (Anthropic)
Only in system cards
Models work on tasks where grading is disclosed clearly, ambiguously or barely. An autorater flags transcripts that reason about the grader.
Reported separately for three disclosure levels with no overall rate.
Made by Anthropic Released Apr 2026 Deception and alignment Tasks kept private Last in a system card Oct 2026
Last model added Oct 2026
- Level
- Expert Specialists with years of training, often over hours of work
- Scored
- Graded by a model A language model scores each response, usually against a rubric.
- Answers
- Agent task Many steps in a terminal, browser, desktop, app, game or robot.
3
System cards report it, from 1 lab
No scores here yet
No lab, leaderboard or Epoch AI run has reported a score for Verbalized grader awareness (Anthropic) yet.
Labs report it only in system cards
We read the headline results of 479 launch posts from 51 labs, and none of them list this benchmark. Some system cards do, below.
3 system cards report it, from 1 lab
Besides launch posts, we read the system cards, model cards and technical reports labs publish with a release. They report far more, including safety and frontier-risk tests.
| Date | Document | Score | As the document writes it | Section |
|---|---|---|---|---|
| 7 Oct 2026 | Claude Haiku 5.5 system card | Verbalized grader awareness by disclosure level: clearly 88.3 (166/188), ambiguously 53.7 (80/149), barely 4.3 (7/161), no overall (Figure 6.4.1.A) | Monitorability | |
| 28 Sep 2026 | Claude Sonnet 5.5 system card | Verbalized grader awareness, no overall: clearly disclosed 231/275, ambiguously disclosed 364/461, barely or not disclosed 19/163 (6.4.1, Figure 6.4.1.A) | Monitorability | |
| 22 Sep 2026 | Claude Opus 5.5 system card | Verbalized grader awareness by disclosure level, no overall: clearly disclosed 84.0% (273/325), ambiguously disclosed 73.1% (380/520), barely or not disclosed 4.5% (7/155) (6.6.2, Figure 6.6.2.A) | Monitorability |
More in Safety and alignment
- FORTRESS In use Paired adversarial and benign national-security misuse prompts. Rubric-graded risk score on adversarial and over-refusal score on benign prompts. Claude Opus 5.5 5.78%
- StrongREJECT In use Forbidden-prompt jailbreak benchmark with a rubric grader that scores how useful a non-refusal actually is.
Inkling 98.6% - AgentDojo In use Agents do tool-use tasks while tools and data carry prompt injections. Attack success rate counts hijacked runs. Grok 4 Fast 0%
- MASK In use Measures whether a model contradicts its own elicited beliefs under pressure to lie. Honesty separated from accuracy.
Muse Spark Open 98.1% - Gray Swan IPI Arena benchmark In use Indirect prompt-injection attacks from public red-teaming competitions replayed on tool, coding and browser agents. Attack success rate at k attempts. GPT-6 Astra 8.50%
- Agentic misalignment suite (Anthropic) In use Agents in simulated company settings face threats or goal conflicts that invite blackmail, leaking or sabotage. Score is the misaligned action rate. Claude Sonnet 4.5 0%