Noise Floor
A statistical tool for measuring the free score a benchmark itself yields, obtained by running the same task set with random or trivial baseline policies to establish a lower bound, used to separate genuine signal from benchmark noise.
Timeline (last 90 days)
The noise floor is the score obtained by running a completely random, trivial or no-op policy on the same task set, representing the lower bound any system can achieve without real capability.
In a binary classification task, if the positive-to-negative sample ratio is 6:4, then a baseline that always outputs "positive" scores 60 points, and an agent scoring 65 has only 5 points of real signal.
All Facts (2)
The noise floor is the score obtained by running a completely random, trivial or no-op policy on the same task set, representing the lower bound any system can achieve without real capability.
50%UnverifiedIn a binary classification task, if the positive-to-negative sample ratio is 6:4, then a baseline that always outputs "positive" scores 60 points, and an agent scoring 65 has only 5 points of real signal.
50%