SPADE Zero Shot - EvoBench Overall 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: Llama 3.1 8B + Llama 3.1 8B InstructEvaluation run: frozen two-checkpoint detector86.4
2SPADE Binoculars: Llama 3.1 8B + Llama 3.1 8B InstructEvaluation run: frozen two-checkpoint detector86
3SPADE Binoculars: Falcon-7B + Falcon-7B-InstructEvaluation run: frozen two-checkpoint detector82.3
4SPADE Fast-DetectGPT: Falcon-7B + Falcon-7B-InstructEvaluation run: frozen two-checkpoint detector82.1
5Qwen3-8B-Base (Harness: SPADE zero-shot Likelihood)82
6Qwen3-8B-Base (Harness: SPADE zero-shot DNA-GPT)80.4
7Qwen3-8B-Base (Harness: SPADE zero-shot Entropy)75.5
8Llama 3.1 8B (Harness: SPADE Zero Shot Likelihood)75.4
9Llama 3.1 8B (Harness: SPADE Zero Shot DNA GPT)75.2
10falcon-7B (Harness: SPADE zero-shot DNA-GPT)74
11SPADE Fast-DetectGPT: Qwen3-8B-Base + Qwen3-8BEvaluation run: frozen two-checkpoint detector73.6
12SPADE Binoculars: Qwen3-8B-Base + Qwen3-8BEvaluation run: frozen two-checkpoint detector73.2
13falcon-7B (Harness: SPADE zero-shot Likelihood)72
14falcon-7B (Harness: SPADE zero-shot NTS)71.3
15Llama 3.1 8B (Harness: SPADE Zero Shot Entropy)63.4

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