SMH-Bench (Interactive Agent) - Compositional Control: leaderboard
Metric: Success rate (%) on Compositional Control (TC2, 250 multi-device, state-dependent, room-dependent and selection-based operations); Environment-Interactive Agent: a ReAct loop that starts from partial observability and queries and controls the HomeEnv simulator through tools under a call budget; final home state checked against the goal with untouched devices preserved; clarification and query answers judged by GPT-5; higher is better. Source: arxiv.org. Saturation forecast: Around December 2026. 13 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 84.4 |
| 2 | Claude Sonnet 4.6 | 81.2 |
| 3 | GPT-5.4 | 78 |
| 4 | Claude Haiku 4.5 | 73.6 |
| 5 | Qwen 3.5 397B A17B | 68.8 |
| 6 | GPT-5.4 Mini | 68.8 |
| 7 | GLM-5 | 65.6 |
| 8 | Qwen 3.5 Plus | 57.2 |
| 9 | Qwen 3.5 9B | 56 |
| 10 | DeepSeek V3.2 (Thinking) | 52.8 |
| 11 | Qwen 3.5 4B | 45.6 |
| 12 | MiniMax-M2.7 | 23.6 |
Interactive version: theaggregate.ai/benchmark?slug=smh-bench-interactive-agent-compositional-control · How It Works · Data refreshed daily, snapshot 2026-09-29.