PolicyShiftBench - Policy Shift Score: leaderboard

Metric: Policy Shift Score (%; over groups of the same image and risk category, the share of label-flipping policy pairs where the model is right on both the permissive and the strict policy, macro-averaged over groups; policy-conditioned image guardrailing: decide whether an image violates the supplied runtime policy bundle (one policy variant per risk category, seven categories); mean of the Adaptive split (policy families seen in training) and the Shift split (held-out policies), 2,000 instances; mean of three evaluations). Source: arxiv.org. Saturation forecast: Around December 2026. 19 models tracked.

Top models

#ModelScore
1Gemini 3 Flash (Preview)50.6
2GPT-5.445.5
3Qwen 3.5 35B A3B (Thinking)29.8
4Claude Sonnet 4.618.8
5Qwen 3.5 35B A3B (Non-reasoning)18.6
6Qwen 3.5 4B (Thinking)17.8
7Qwen 2.5 VL 7B4.8
8Qwen 3.5 4B (Non-reasoning)4.4
9Qwen 3.5 0.8B (Non-reasoning)1.8
10Qwen 3.5 2B (Non-reasoning)0

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