Deep personalization boosts on-device agent timing macro-F1 by 2.4–3.0x
EchoShao8899 · x · 2026-11-17
On OSWorld-Persona, the author shows Deep Personalization improves when-to-help timing macro-F1 by 2.4–3.0×: Qwen-3.5-9B goes 0.186 → 0.483, Qwen-3.5-35B-A3B 0.183 → 0.433, and Nemotron-3-Nano-Omni 0.118 → 0.349. Across 10 real users with varied AI familiarity, Coco's macro-F1 proxy rose from 0.199 to 0.515 in 7 days, as learned preferences evolved from broad cues to workflow-specific decisions. Contextual memory built by the on-device Observer also improves timing judgments.
Related event: Coco: An On-Device Desktop Agent That Learns When to Help(9 posts)→
More from coding & agent
- Bria open-sources RMBG 3.0 background removal model, ships MCP server and Claude skill — victormustar · 2026-11-17
- 6 AI Agents Design an Apartment's Full MEP: 4,871 Layouts, 11 Clashes Fixed — minchoi · 2026-11-17
- Quick-start guides in 2026: just show me the prompt, not npm install — MxMnr · 2026-11-17
- W&B Sweeps adds Optuna and Ax: pick your HPO engine with one line of YAML — coreweavedevs · 2026-11-17
- How do you find which step blew your LLM budget? Devs discuss per-step cost tracking — Youthify · 2026-11-17
- Gemini Managed Agents Adds Self-Hosted Sandboxes: Google Runs the Loop, You Run the Tools — _philschmid · 2026-11-17