SPADE Zero Shot - RAID Abstracts 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 Binoculars: Llama 3.1 8B + Llama 3.1 8B InstructEvaluation run: frozen two-checkpoint detector92.9
2falcon-7B (Harness: SPADE zero-shot NTS)92.4
3SPADE Fast-DetectGPT: Llama 3.1 8B + Llama 3.1 8B InstructEvaluation run: frozen two-checkpoint detector91.1
4SPADE Binoculars: Falcon-7B + Falcon-7B-InstructEvaluation run: frozen two-checkpoint detector90.2
5Llama 3.1 8B (Harness: SPADE Zero Shot NTS)89.2
6SPADE Fast-DetectGPT: Falcon-7B + Falcon-7B-InstructEvaluation run: frozen two-checkpoint detector87.6
7falcon-7B (Harness: SPADE zero-shot DNA-GPT)82.2
8SPADE Binoculars: Qwen3-8B-Base + Qwen3-8BEvaluation run: frozen two-checkpoint detector80.7
9SPADE Fast-DetectGPT: Qwen3-8B-Base + Qwen3-8BEvaluation run: frozen two-checkpoint detector76.6
10Qwen3-8B-Base (Harness: SPADE zero-shot NTS)76.3
11Qwen3-8B-Base (Harness: SPADE zero-shot Likelihood)75.5
12Llama 3.1 8B (Harness: SPADE Zero Shot Likelihood)73.7
13Llama 3.1 8B (Harness: SPADE Zero Shot DNA GPT)73.1
14falcon-7B (Harness: SPADE zero-shot Likelihood)72.2
15Qwen3-8B-Base (Harness: SPADE zero-shot DNA-GPT)71

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