SPADE Zero Shot - EvoBench PubMed AUROC: leaderboard

The SPADE paper compares six zero-shot AI-text detectors using three released language-model backbones. AUROC and true-positive rate at 5% false positives remain separate. Paired base/instruct checkpoints are retained for Binoculars and Fast-DetectGPT. Learned soft prompts and supervised detector weights are distinct artifacts. This board reports aUROC (%). Higher is better.

Metric: AUROC (%). Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 18 models tracked. Updated 14 Nov 2026.

Top models

#ModelSettingScoreOverall rank
1SPADE Fast-DetectGPT: Falcon-7B + Falcon-7B-InstructEvaluation run: frozen two-checkpoint detector91.5
2SPADE Binoculars: Falcon-7B + Falcon-7B-InstructEvaluation run: frozen two-checkpoint detector91.2
3Qwen3-8B-Base (Harness: SPADE zero-shot Likelihood)90.8
4Qwen3-8B-Base (Harness: SPADE zero-shot Entropy)90.7
5Qwen3-8B-Base (Harness: SPADE zero-shot DNA-GPT)87.5
6SPADE Fast-DetectGPT: Llama 3.1 8B + Llama 3.1 8B InstructEvaluation run: frozen two-checkpoint detector87.4
7SPADE Binoculars: Llama 3.1 8B + Llama 3.1 8B InstructEvaluation run: frozen two-checkpoint detector86.9
8Llama 3.1 8B (Harness: SPADE Zero Shot Likelihood)82.8
9falcon-7B (Harness: SPADE zero-shot Likelihood)82.7
10Llama 3.1 8B (Harness: SPADE Zero Shot DNA GPT)82.2
11falcon-7B (Harness: SPADE zero-shot DNA-GPT)81.3
12Llama 3.1 8B (Harness: SPADE Zero Shot Entropy)76.7
13SPADE Fast-DetectGPT: Qwen3-8B-Base + Qwen3-8BEvaluation run: frozen two-checkpoint detector71.3
14SPADE Binoculars: Qwen3-8B-Base + Qwen3-8BEvaluation run: frozen two-checkpoint detector69.1
15falcon-7B (Harness: SPADE zero-shot Entropy)67.9

Interactive version: theaggregate.ai/benchmark?slug=spade-zero-shot-evobench-pubmed-auroc · How It Works · Data refreshed daily, snapshot 2026-11-16.