RMGAP - Listwise LLM Judge - Best-of-4: leaderboard
RMGAP (Zhou and Li, 2026) tests whether reward models follow a preference stated in the prompt: each of 1,097 instances pairs four responses in deliberately different styles with prompts that each ask for one of those styles (three paraphrases each), across chat, writing, reasoning and safety. The judge model ranks the four shuffled responses in one call and ranks map to scores. Best-of-4 accuracy is the percentage of cases in which the requested response scores above all three alternatives, averaged over the four domains; higher is better.
Metric: Best-of-4 accuracy (%; chosen response scored above all three alternatives, mean of chat, writing, reasoning and safety; the model ranks the four shuffled responses in one call, ranks mapped to scores 4-1). Source: arxiv.org. Saturation forecast: Around February 2027. 5 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3 235B A22B 2507 | 49.27 | #613 |
| 2 | DeepSeek V3.2 | 46.75 | #403 |
| 3 | Qwen 2.5 7B Instruct | 36.84 | #2583 |
| 4 | Qwen 3 8B | 33.68 | #1838 |
| 5 | Llama 3.1 8B Instruct | 24.88 | #4277 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=rmgap-listwise-llm-judge-best-of-4 · How It Works · Data refreshed daily, snapshot 2026-11-12.