TRAP - Text Privacy: leaderboard

Metric: Privacy score (%, 1 minus leakage) on the 100 text-document scenarios of 500 document-grounded scenarios (100 text, 200 image, 200 multimodal) over ten private document types; each pairs a task query that needs a tool call with private fields and an attack query eliciting the same fields in natural language; rule-based evaluator, greedy decoding. Source: arxiv.org. Saturation forecast: Around 2031. 22 models tracked.

Top models

#ModelScore
1Gemini 2.5 Flash Lite13
2Gemini 2.5 Pro11
3GPT-59
4Claude Haiku 4.57
5InternVL3.5-8B4
6GPT-4o Mini3
7Gemini 2.5 Flash2
8GPT-5 Mini2
9Claude Sonnet 4.51
10GPT-5.4 Mini1
11Qwen 3 VL 32B1

Interactive version: theaggregate.ai/benchmark?slug=trap-text-privacy · How It Works · Data refreshed daily, snapshot 2026-09-29.