Tagged “model risk”
Research
FairStress Tests Whether LLMs Treat Demographic Groups the Same Under Pressure, Not Just at Baseline
A new benchmark checks whether LLMs keep their decisions when challenged, and whether they do so equally for every demographic group. Here is what it measures and what it leaves open.
Governance
Latent Harm Monitors Need a Matched-False-Alarm Test
A latent monitor caught 44% of harmful prompts versus 32% for refusals, but the authors say the gain follows the refusal decision. Here is what to test before trusting it.