EYT-Bench (PersonaMem-v2) - Emotion: leaderboard

Metric: Per-turn emotion accuracy (%; the target model first predicts the user emotion label from 15 classes, scored against the simulator gold label, before generating its reply; multi-turn dialogues of up to 10 turns with a persona-grounded Gemini 3.1 Pro user simulator and a Gemini 3.1 Pro judge (reasoning effort high, temperature 0); PersonaMem-v2 pool (100 dialogues; personas are free-text paragraphs distilled from long real user-assistant interactions)). Source: arxiv.org. Saturation forecast: Around December 2026. 17 models tracked.

Top models

#ModelScore
1DeepSeek V4 Pro75.3
2DeepSeek V4 Flash64.5
3GPT-5.550
4Gemini 3.1 Pro (Preview) (Thinking)42.2
5Claude Sonnet 4.623
6Claude Opus 4.719.8
7Qwen 3.5 397B A17B13.3
8Qwen 3.5 27B13.1
9Seed 2.0 Mini12.8
10Seed 2.0 Pro12.7
11Qwen 3.5 35B A3B12.4
12Seed 2.0 Lite9.4

Interactive version: theaggregate.ai/benchmark?slug=eyt-bench-personamem-v2-emotion · How It Works · Data refreshed daily, snapshot 2026-09-29.