Gadi Singer
Titles as the sources gave them. In 1 entry, with 0 checked quotations.
In the Base
Every entry they appear in.
BMMS: Policy-Conditioned Measurement of Behavioral Compliance in Language Models
A preprint introduces the Behavioral Misalignment Metric Suite (BMMS), which takes a policy corpus and returns a compliance profile showing whether a language model system does what a given rulebook requires. It is tested on 1,003 obligations from eleven documents.