Neo-Classic - Parallel Couplet Matching: leaderboard

Metric: Accuracy (%; select the matching line of a parallel middle couplet of a contemporary regulated verse, four options; accuracy on a 100-question sample of the task, as reported in the paper's full results table (accuracy@IF); temperature 0.1, top-p 0.9). Source: arxiv.org. Saturation forecast: Around 2029. 8 models tracked.

Top models

#ModelScore
1DeepSeek V3.272.1
2Qwen 3 Max72
3Gemini 3 Pro (Preview)68
4Yi 1.5 34B Chat65
5Gemini 2.5 Pro61
6GPT-4o57.2

Interactive version: theaggregate.ai/benchmark?slug=neo-classic-parallel-couplet-matching · How It Works · Data refreshed daily, snapshot 2026-09-26.