Polar (US): leaderboard
Polar (2026) measures political bias in language models from option likelihoods on multiple-choice items built from real political statements, on economic and sociocultural axes. This board is the ICAT score from 0 to 100 on the U.S. set in English, combining how often the unrelated continuation is rejected with how balanced the model is between the two political continuations, averaged over both axes; higher is better.
Metric: Total ICAT (0-100), the StereoSet idealized context association score adapted to political bias: the language-model score LMS (share of items where the unrelated continuation is not chosen) times the neutrality score NS (how close the preference between the left or progressive and the right or conservative continuation is to balance, averaged over issue categories); options chosen by length-normalized log-likelihood in LM Evaluation Harness; higher means more balanced and competent; mean of the economic and sociocultural axes, U.S. political statements in English. Source: arxiv.org. 38 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Llama 3.2 1B | 79.27 | #9258 |
| 2 | Qwen 3 0.6B | 78.72 | #8342 |
| 3 | Qwen3-0.6B-Base | 78.69 | #8156 |
| 4 | Qwen3-8B-Base | 77.8 | #1721 |
| 5 | Qwen3-1.7B-Base | 77.75 | #4890 |
| 6 | Llama 3.2 1B Instruct | 76.97 | #9047 |
| 7 | Llama 3.2 3B | 76.91 | #7830 |
| 8 | Qwen 3 8B | 76.15 | #1830 |
| 9 | Midm-2.0-Base-Instruct | 76.12 | #1921 |
| 10 | Llama 3.2 3B Instruct | 76.1 | #7147 |
| 11 | Qwen 3 1.7B | 76.03 | #5520 |
| 12 | Qwen3-4B-Base | 75.97 | #2515 |
| 13 | Qwen 3 4B | 75.36 | #2281 |
| 14 | Qwen3-14B-Base | 75.14 | #1504 |
| 15 | Mistral 7B Instruct (v0.3) | 74.89 | #6653 |
Interactive version: theaggregate.ai/benchmark?slug=polar-us · How It Works · Data refreshed daily, snapshot 2026-11-13.