KnowU-Bench - Personalized (Easy): leaderboard

Personalized mobile assistance on KnowU-Bench (2026), easy split: 43 vague requests such as ordering lunch whose correct completion depends on a hidden user profile, which the agent must infer from noisy behavior logs or by asking an LLM-simulated user. Success rate in percent; higher is better.

Metric: Success rate (%) on the 43 easy personalized tasks (vague instructions resolved from the user logs or by asking the simulated user) 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

#ModelScoreOverall rank
1Claude Sonnet 4.644.2#173
2Gemini 3.1 Pro (Preview)34.9#99
3Seed 2.0 Pro32.6#222
4Qwen 3.5 122B A10B30.2#346
5Qwen 3.5 397B A17B25.6#264
6Qwen 3 VL 32B18.6#622
7UI-Venus-1.5-8B18.6
8MAI-UI-8B16.3
9Qwen 3.5 9B9.3#828
10GUI-OWL-1.5-8B9.3
11Qwen 3 VL 8B7#917

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-easy · How It Works · Data refreshed daily, snapshot 2026-11-21.