SMH-Bench (Interactive Agent) - Ambiguous Intent: leaderboard

Metric: Success rate (%) on Ambiguous Intent (TC3, 150 subjective complaints, scene-level goals and slot-missing requests, acting or asking for clarification as appropriate); 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: Estimated already saturated. 13 models tracked.

Top models

#ModelScore
1Gemini 3.1 Pro (Preview)92
2Claude Sonnet 4.691.3
3Qwen 3.5 397B A17B86.7
4GPT-5.4 Mini86
5Claude Haiku 4.585.3
6DeepSeek V3.2 (Thinking)85.3
7GPT-5.484.7
8GLM-583.3
9Qwen 3.5 Plus76.7
10Qwen 3.5 4B72
11Qwen 3.5 9B69.3
12MiniMax-M2.752

Interactive version: theaggregate.ai/benchmark?slug=smh-bench-interactive-agent-ambiguous-intent · How It Works · Data refreshed daily, snapshot 2026-09-29.