Model techniques map
Techniquesevaluationhuman & real-world evaluation

specific method · filed under evaluation

Held-out evaluation

Reserves benchmarks from development decisions and evaluates them only for final validation.

Also called held-out generalization gates.

source
1
model
1
lab adopt it
1
strongest
default

How sources treat it

One count per evidence span, weakest treatment to strongest.

default 1

Documented in

Evidence

1 span quoted from the sources, strongest treatment first.

These benchmarks were reserved for final validation: they were not used for training-time monitoring, checkpoint selection, or any other development decisions, and were evaluated only once after the final model was produced.

defaultevaluation onlyin Nemotron 3 UltraNVIDIA

Filed alongside

Other methods under evaluation :: human & real-world evaluation.