PERCEIVE - Personalized Emotion (Twitter): leaderboard
Personalized reader emotion prediction (the same task given the reader profile) on PERCEIVE (2026), reader-centric social-media data that links posts with real reader comments, behaviours and profiles, English Twitter test split, with the model prompted directly. Macro-F1 times 100 (0-100); higher is better.
Metric: Personalized emotion prediction (Task B: Task A with the reader profile), macro-F1 over the classes (times 100, so 0-100) on the held-out test split (8:1:1 partition) of PERCEIVE, reader-centric social-media data with real reader comments, behaviours and profiles, English Twitter subset; the model is prompted directly; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 3 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | PERCEIVE Qwen3-plus (checkpoint unspecified) | 54.56 | |
| 2 | DeepSeek V3.2 | 52.26 | #419 |
| 3 | PERCEIVE GLM4.5-plus (checkpoint unspecified) | 49.07 |
No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=perceive-personalized-emotion-twitter · How It Works · Data refreshed daily, snapshot 2026-11-21.