SciFigPlag-Bench: leaderboard

Metric: Overall score (%; mean of pairwise-detection accuracy, source-attribution accuracy, reuse-type joint accuracy and paired reuse-localization accuracy; SciFigPlag-Bench pairs of source and suspicious scientific figures (2,582 plagiarized pairs from documented real cases and taxonomy-guided synthesis, 2,541 visually similar negatives); the same task-specific prompts for every model, temperature 0). Source: arxiv.org. Saturation forecast: Around December 2026. 16 models tracked.

Top models

#ModelScore
1Gemini 3 Flash83
2Gemma 4 31B (IT)82.7
3Qwen 3.6 35B A3B80.5
4GPT-5.480.1
5Qwen 3.5 35B A3B79.8
6Claude Sonnet 4.677.6
7Gemma 4 26B A4B (IT)53.5
8InternVL3.5-8B45.8
9Mistral Small 437.3
10Pixtral-12B-240933.1
11Qwen 3.5 2B30.9
12Gemma 3 4B (IT)28.6

Interactive version: theaggregate.ai/benchmark?slug=scifigplag-bench · How It Works · Data refreshed daily, snapshot 2026-09-29.