Personalized RewardBench - Arts and Entertainment: leaderboard

Personalized preference judgment on the 767 Arts and Entertainment test pairs of Personalized RewardBench (2026): for LaMP-QA queries, the chosen response follows the user-specific rubric aspects while the rejected one, of similar general quality, does not. Pairwise accuracy in percent (scalar models must score the chosen response higher, generative models pick it); higher is better.

Metric: Pairwise accuracy (%) on the 767 Arts and Entertainment test pairs of Personalized RewardBench, where the chosen response follows the user-specific rubric aspects and the rejected one avoids them while keeping general quality (human-validated); scalar models must score the chosen response strictly higher, generative models pick one of the two in shuffled order; no user profile in the input; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 17 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Gemini 3 Flash72.36#175
2R3-Qwen3-8B-14k67.8
3Claude Sonnet 4.667.28#173
4R3-Qwen3-14B-14k66.88
5Skywork-Reward-V2-Llama-3.1-8B66.62
6InternLM2-7B-Reward65.97
7GPT-5.165.45#240
8RM-R1-Qwen2.5-Instruct-7B65.06
9RM-R1-Qwen2.5-Instruct-14B64.8
10mR3-Qwen3-8B64.28
11InternLM2-20B-Reward63.23
12R3-Qwen3-4B-14k62.71
13Skywork-Reward-V2-Llama-3.2-1B62.06
14mR3-Qwen3-4B61.54
15Skywork-Reward-V2-Llama-3.2-3B60.89

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=personalized-rewardbench-arts-and-entertainment · How It Works · Data refreshed daily, snapshot 2026-11-21.