PreAct-Bench - Politics: leaderboard

Metric: Prefix Foresight F1 with the full prefix (PFF(100), binary F1 of the unethical class, 0-1 scaled to %): given an agent trajectory truncated just before its first overt unethical step, the monitor predicts whether the trajectory will turn unethical; paired ethical and unethical trajectories per task; default generation parameters; higher is better; the politics domain. Source: arxiv.org. Saturation forecast: Around April 2028. 18 models tracked.

Top models

#ModelScore
1Gemma 3 27B82.1
2Qwen 3 30B A3B 2507 Instruct81.3
3GPT-5.477.7
4Qwen 3 8B (Non-reasoning)74.5
5Qwen 3 8B (Thinking)74.3
6Gemma 3 4B66.7
7Qwen 3 0.6B (Non-reasoning)66.7
8GPT-4o62.1
9Qwen 3 30B A3B 2507 (Thinking)60.9
10Qwen 3 0.6B (Thinking)58.8
11Qwen 3 4B 2507 (Thinking)55.4
12Qwen 3 4B 2507 Instruct54
13GPT-OSS-20B53.5
14GPT-OSS-120B51.7

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