Benchmark · RAID+ · scorer 2.7.2

Not all machine writing is equally generic.

We scored 7,627 machine-generated abstracts. Llama 3.3 70B produced about 4.1× DeepSeek V3’s rate of text at or above the generic gate.

Results

ModelTextsMean scoreMedianAt or above 25
DeepSeek V31,99514.59.510.1%
Gemini 3.1 Pro1,998179.518.2%
Gemma 3 27B1,63421.618.430.4%
Llama 3.3 70B2,00025.521.341.7%
Overall7,62719.615.624.9%

What this is not

RAID+ records which model produced a text, not whether the text is any good. These are score distributions, not editorial quality, not authorship accuracy, and not a comparison with human writing. The writing score measures surface style — stock phrasing, mechanical rhythm, dense passages, distracting formatting — and it does not identify who wrote anything.

Every text here is an abstract. Academic abstracts are a formal register that the scorer treats carefully, and results on other kinds of writing will differ. A separate corpus of samples the scorer must not flag is kept in the repository for exactly this reason.

Method

RAID+ is an MIT-licensed extension of the peer-reviewed RAID benchmark. We scored the 8,000 pinned rows at revision 317600a0c8dd; 7,627 remained after excluding 373 failed or empty generations. The generic gate is 25.

Reproduce it, or disagree with it

Everything needed to rerun this is public and MIT-licensed. If you build a competing tool, run it against the same pinned corpus and publish the numbers — a comparison nobody controls is worth more than any claim we make here.

Download the results as CSV · Benchmark runner and full method · Star the repository · Run the score in CI

Cite this benchmark

Every run is pinned to a scorer build, so a citation names a version rather than “the site as of today”. This release is 2.7.2, scorer sha256:99217fff003a.

Mishra, M. Zero Slop writing scores over RAID+ (version 2.7.2).
  7,627 generations, scorer sha256:99217fff003a.
  https://zero-slop.ai/benchmark/
@misc{zeroslop_raidplus_2_7_2,
  author = {Mishra, Manav},
  title  = {Zero Slop writing scores over {RAID+}},
  year   = {2026},
  note   = {Version 2.7.2, 7,627 generations},
  url    = {https://zero-slop.ai/benchmark/}
}