VMAF-NEG, the No Enhancement Gain variant of VMAF, is a model mode that stops the score from rewarding image enhancement. Because standard VMAF was trained to predict what looks good, sharpening or boosting contrast before encoding can raise the score even though nothing about the compression improved — the encoder has simply learned to please the metric. That behaviour is fine when you genuinely enhanced the picture for viewers, but it corrupts an encoder-versus-encoder test, where the goal is to measure compression gain alone. VMAF-NEG disables the enhancement-driven boost so a sharpening trick cannot inflate the result; the rule of thumb is to use the standard model for predicting delivered quality and NEG when comparing encoders so nobody can game the comparison. It addresses the broader hazard that any objective metric can be optimized past the point where it tracks the eye, and in the 2026 VMAF v1 generation the enhancement-gain limit is turned on by default.
