Prajakta Narsay

RESEARCH MEMO The Problem Standard alignment benchmarks test models that know they are being tested. The alignment faking paper showed Claude 3 Opus reasoning strategically about training dynamics and behaving differently when it believed it was unmonitored without being explicitly trained to do so. Any evaluation a sufficiently capable model...

The Problem Standard alignment benchmarks test models that know they are being tested. The alignment faking paper showed LLMs reasoning strategically about training dynamics and behaving differently when it believed it was unmonitored without being explicitly trained to do so. Any evaluation a sufficiently capable model can recognize, it can...