Discoverable memorization of training data
Only in system cards
Models continue 50-token prefixes sampled from their own training data. The test counts exact and approximate copies of the true suffix.
In-house Google method on over 700,000 training documents. Reports usually plot rates on a log scale without printed numbers.
Made by Google DeepMind Refusing harmful requests Tasks kept private Last in a system card Jul 2025
Last model added Jul 2025
- Level
- Everyday Most adults, without training
- Scored
- Checked automatically A program checks the result against an answer key, unit tests or the expected final state.
- Answers
- Short answer A number, word or short string matched to a key.
2
System cards report it, from 1 lab
No scores here yet
No lab, leaderboard or Epoch AI run has reported a score for Discoverable memorization of training data 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.
2 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 Jul 2025 |  Gemini 2.5 Pro technical report | Total memorization rate, exact and approximate, log-scale chart without printed values; no personal information found in memorized outputs (Figure 8, Section 5.6) | Safety | |
| 7 Jul 2025 | Gemini 2.5 Flash technical report | Total memorization rate, exact and approximate, log-scale chart without printed values; no personal information found in memorized outputs (Figure 8, Section 5.6) | Safety |
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%