GenderBench - DeMET Relationship Conflict Gap: leaderboard
The model judges who is in the right in relationship conflicts from the DeMET scenarios (Levy et al.) with the partners' genders varied; the number is the absolute gap in how often each gender is judged right.
Metric: Gap in how often each gender is judged in the right, absolute (0-1 normalized, lower is better). Source: genderbench.readthedocs.io. 12 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Llama 3.1 8B Instruct | 0.13 | #4246 |
| 2 | Phi-4 | 0.27 | #1913 |
| 3 | Gemini 2.0 Flash Lite | 0.28 | #1032 |
| 4 | Llama 3.3 70B Instruct | 0.29 | #1357 |
| 5 | Gemini 2.0 Flash | 0.31 | #704 |
| 6 | Claude 3.5 Haiku | 0.33 | #1553 |
| 7 | GPT-4o Mini | 0.38 | #1623 |
| 8 | Mistral 7B Instruct (v0.3) | 0.44 | #6653 |
| 9 | Mistral Small 3 | 0.46 | #1777 |
| 10 | GPT-4o | 0.54 | #741 |
| 11 | Gemma 2 9B (IT) | 0.54 | #2262 |
| 12 | Gemma 2 27B (IT) | 0.64 | #1782 |
Interactive version: theaggregate.ai/benchmark?slug=genderbench-demet-relationship-conflict-gap · How It Works · Data refreshed daily, snapshot 2026-11-13.