What Does This Mean?
Any evaluation or benchmark run of a model conducted with its safety guardrails intentionally turned off, used as the general umbrella term for this practice — of which a 'safety-off run' is the specific research-team procedure for probing raw model capability.
REAL-WORLD EXAMPLE
"Any safety-disabled eval needs to happen in an isolated environment, no exceptions, regardless of who's running it or why."
LORE & ORIGIN
Named directly and plainly as a broader category label for evaluations run without active safety mechanisms, covering both deliberate capability research and less controlled contexts where safety was disabled for other reasons.
MENTIONS OVER TIME
Radar observation data not yet available for this term.
RELATED SEARCHES
TOP PLATFORMS
Not enough Radar data yet to show a platform breakdown.
WHEN DID YOU FIRST HEAR THIS?
CLICK AN OPTION BELOW TO CAST YOUR VOTE.
[ VOTE TO REVEAL COMMUNITY RESULTS ]