PriVE-Bench - Industry Attributes: leaderboard

Metric: Counterfactual accuracy (%) in the industry attributes domain; paired original and counterfactual images in five domains, counterfactual-only input, answers classified by a text-only judge (GPT-4o-mini for closed models, Qwen3-8B for open models); temperature 0, reasoning off or minimal. Source: arxiv.org. Saturation forecast: Around December 2026. 8 models tracked.

Top models

#ModelScore
1GPT-5 (Minimal)85.1
2Claude Sonnet 467.3
3InternVL3.5-8B56.4
4Qwen 3 VL 32B Instruct53.6
5Qwen 3 VL 8B Instruct51.7
6Gemma 3 12B (IT)29.7
7Gemma 3 27B (IT)21.6

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