Skip to content

Evaluation

Adversarial long-context needle-in-a-haystack safety evaluation

Adversarial version of text needle-in-a-haystack: adversarial needle (T2T policy-violating prompt) inserted in Paul Graham essay haystack across 25 settings (1k-1M tokens, varying depth); automated short-context T2T evaluation applied

Harmful content & safety Google DeepMind — Gemini

How it was run

Open a release to see the setup, scoring, and other details that the lab published. Only fields the lab actually disclosed are shown.

Gemini 1.5
Google DeepMind — Gemini

rich detail 8 method fields published

See how this test was run