KnowU-Bench - Personalized Score (Hard): leaderboard
Partial-credit score for personalized mobile assistance on KnowU-Bench (2026), hard split: 43 vague requests at the higher of the two difficulty levels, each scored from rule checks fused with a rubric-based judgment of preference alignment. Mean task score in percent; higher is better.
Metric: Average score (%, the 0-1 task score times 100; rule checks fused with a rubric judge) on the 43 hard personalized tasks in the reproducible KnowU-Bench Android emulator, noisy full-history memory setting (the complete user logs plus irrelevant history), gpt-4o as the user simulator grounded in a hidden user profile, scored by rule-based checks fused with a rubric-conditioned LLM judge; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 11 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Sonnet 4.6 | 80 | #173 |
| 2 | Gemini 3.1 Pro (Preview) | 75 | #99 |
| 3 | Qwen 3.5 122B A10B | 60 | #346 |
| 4 | Seed 2.0 Pro | 57 | #222 |
| 5 | Qwen 3.5 397B A17B | 48 | #264 |
| 6 | UI-Venus-1.5-8B | 40 | |
| 7 | GUI-OWL-1.5-8B | 34 | |
| 8 | MAI-UI-8B | 31 | |
| 9 | Qwen 3 VL 32B | 26 | #622 |
| 10 | Qwen 3 VL 8B | 25 | #917 |
| 11 | Qwen 3.5 9B | 18 | #828 |
No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=knowu-bench-personalized-score-hard · How It Works · Data refreshed daily, snapshot 2026-11-21.