WatchAct - Nonverbal Cue: leaderboard

Metric: Plan success rate (%; the VLM's high-level Pick/Place/Open/Close plan, predicted from a real human video and an instruction, run through a symbolic executor and checked against the task's success predicate; 195 instances). Source: arxiv.org. Saturation forecast: Around February 2028. 9 models tracked.

Top models

#ModelScore
1Gemini 3 Flash23.1
2GPT-5.421
3Gemini 3.1 Pro (Preview)18
4Qwen 3 VL 235B A22B Instruct6.7
5Qwen 3 VL 235B A22B (Thinking)6.2
6Gemini 3.1 Flash Lite5.1
7GPT-5.14.6

Interactive version: theaggregate.ai/benchmark?slug=watchact-nonverbal-cue · How It Works · Data refreshed daily, snapshot 2026-09-26.