Grading Accuracy You Can Trust

Our scoring engine powered by DeepSeek V3 delivers expert-level accuracy at machine speed -- with detailed, actionable feedback on every essay.

QWK 0.77
Quadratic Weighted Kappa -- exceeds average human-to-human agreement
94%
Within 1 point of expert human graders
<5s
Scoring speed per essay
24,728
Essays tested in our benchmark

Accuracy Comparison

TeachShieldQWK 0.77
Human-to-Human AgreementQWK 0.60-0.75
GPT-4 Zero-ShotQWK 0.45-0.55
ChatGPT Zero-ShotQWK 0.25-0.35

The Scoring Engine

DeepSeek V3 Scoring

Powered by DeepSeek V3, our engine scores essays and generates detailed, personalized feedback -- quoting student work and suggesting specific improvements.

Benchmark Validated

QWK 0.77 achieved in benchmark testing on 24,728 expert-graded essays (ASAP 2.0 dataset). Benchmark results reflect testing conditions and are not guaranteed in all production scenarios.

What is QWK?

Quadratic Weighted Kappa (QWK) is the standard metric for evaluating automated essay scoring systems. It measures agreement between the automated scores and expert human graders, penalizing larger disagreements more heavily than smaller ones.

  • QWK 1.0 = perfect agreement with human graders
  • QWK 0.60-0.75 = typical human-to-human agreement range
  • QWK 0.77 = TeachShield's score, exceeding average human agreement

Ready to grade smarter?

Try it free -- 50 essays per month, no credit card required.

Start Free Today