EgoGapBench - Indirect Actions: leaderboard

Metric: Accuracy (%; zero-shot multiple choice on 350 indirect-action single-image items built from COCO multi-agent scenes: choose the action available from the given viewpoint, with distractors copying the singleton action of another visible agent (the correct answer is a passive response such as watching or waiting); predicted option string-matched to the key; chance about 30%). Source: arxiv.org. Saturation forecast: Around December 2026. 9 models tracked.

Top models

#ModelScore
1GPT-5.4 (Medium)94.3
2GPT-5.4 (Low)90.6
3gemma-4-E4B-it84.6
4Gemini 3 Flash (Preview) (Low)60.3
5GPT-4o (2024-11-20)52.3
6Qwen 3.5 9B45.7
7Qwen 2.5 VL 7B Instruct41.1
8InternVL3.5-8B30.9

Interactive version: theaggregate.ai/benchmark?slug=egogapbench-indirect-actions · How It Works · Data refreshed daily, snapshot 2026-09-29.