PEC-Home (RAG) - Dynamic Preferences Level 2: leaderboard

Metric: Execution accuracy (%; 712 smart-home commands at ellipsis level 2 of 4 (progressively elliptical), functionally correct device operations; dialogue history retrieved from a vector store, top-3). Source: arxiv.org. Saturation forecast: Estimated already saturated. 10 models tracked.

Top models

#ModelScore
1Qwen 2.5 7B Instruct91.57
2Gemma 2 27B (IT)88.37
3GPT-4o88.34
4Qwen 2.5 32B Instruct79.92
5Gemma 2 9B (IT)79.78
6Qwen 2.5 72B Instruct79.21
7DeepSeek V378.65
8Mistral 7B Instruct (v0.3)70.93
9Qwen 2.5 14B Instruct66.57
10Llama 3.1 8B Instruct58.29

Interactive version: theaggregate.ai/benchmark?slug=pec-home-rag-dynamic-preferences-level-2 · How It Works · Data refreshed daily, snapshot 2026-09-26.