Benchmark

SHADE-Arena

An evaluation benchmark for assessing AI deception detection capability. It measures a classifier's ability to distinguish conversations containing deceptive behavior from those without deception, commonly using AUC as the core metric.

Timeline (last 90 days)

Oct 13

In the SHADE-Arena evaluation, the probe reached 98.8% AUC, surpassing the Opus 5.5-based text monitoring baseline.

Unverified50%

All Facts (1)

Unverified

In the SHADE-Arena evaluation, the probe reached 98.8% AUC, surpassing the Opus 5.5-based text monitoring baseline.

50%

Source Articles

Probes Catch AI Lies: How White-Box Detection Unmasks Hidden Model ObjectivesOct 13