SocialPersona - Dialogue (Direct Profile): leaderboard
SocialPersona (2026) tests whether multimodal models can infer what a person cares about from a real social-media timeline (post text, image captions and timestamps) and use it. This board rates personalized recommendations written from a profile the model first built in one pass. Scored 0 to 5 by a GPT-5.5 judge as the mean of interest coverage, concreteness and fluency; higher is better.
Metric: Dialogue score (0 to 5): GPT-5.5 judge mean of interest coverage, concreteness and fluency of responses to stable-interest recommendation and recent-interest exploration requests, judged against the gold profile; the response is written from a profile the model first built from the whole timeline in one pass; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 6 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.4 | 3.55 | #123 |
| 2 | Qwen 3 VL 8B Instruct | 3.39 | #981 |
| 3 | GPT-4o Mini | 3.01 | #1623 |
| 4 | Qwen 3.5 35B A3B | 2.91 | #482 |
| 5 | Gemini 2.5 Flash | 2.68 | #453 |
| 6 | Qwen 2.5 VL 7B Instruct | 2.17 | #1833 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=socialpersona-dialogue-direct-profile · How It Works · Data refreshed daily, snapshot 2026-11-13.