SPADE Zero Shot - EvoBench Writing 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
| # | Model | Setting | Score | Overall rank |
|---|---|---|---|---|
| 1 | SPADE Binoculars: Llama 3.1 8B + Llama 3.1 8B Instruct | Evaluation run: frozen two-checkpoint detector | 98.8 | |
| 2 | SPADE Binoculars: Falcon-7B + Falcon-7B-Instruct | Evaluation run: frozen two-checkpoint detector | 98.7 | |
| 3 | SPADE Fast-DetectGPT: Llama 3.1 8B + Llama 3.1 8B Instruct | Evaluation run: frozen two-checkpoint detector | 98.5 | |
| 4 | SPADE Fast-DetectGPT: Falcon-7B + Falcon-7B-Instruct | Evaluation run: frozen two-checkpoint detector | 98.5 | |
| 5 | Qwen3-8B-Base (Harness: SPADE zero-shot Likelihood) | 97.4 | ||
| 6 | Qwen3-8B-Base (Harness: SPADE zero-shot Entropy) | 96.7 | ||
| 7 | falcon-7B (Harness: SPADE zero-shot NTS) | 96.7 | ||
| 8 | Llama 3.1 8B (Harness: SPADE Zero Shot Likelihood) | 94.8 | ||
| 9 | Qwen3-8B-Base (Harness: SPADE zero-shot DNA-GPT) | 90.4 | ||
| 10 | falcon-7B (Harness: SPADE zero-shot DNA-GPT) | 90.1 | ||
| 11 | Llama 3.1 8B (Harness: SPADE Zero Shot Entropy) | 89.4 | ||
| 12 | falcon-7B (Harness: SPADE zero-shot Likelihood) | 89 | ||
| 13 | Llama 3.1 8B (Harness: SPADE Zero Shot DNA GPT) | 87.5 | ||
| 14 | SPADE Binoculars: Qwen3-8B-Base + Qwen3-8B | Evaluation run: frozen two-checkpoint detector | 82.8 | |
| 15 | Llama 3.1 8B (Harness: SPADE Zero Shot NTS) | 81.5 |
Interactive version: theaggregate.ai/benchmark?slug=spade-zero-shot-evobench-writing-auroc · How It Works · Data refreshed daily, snapshot 2026-11-16.