GST-Bench - Top-Down Selection (Medium): leaderboard

Metric: Accuracy (%; multiple choice of the top-down image of the explored scene, medium distractors; zero-shot, greedy decoding, official prompt templates, over human-verified questions on 6,790 minutes of simulator-rendered egocentric exploration video). Source: arxiv.org. Saturation forecast: Around December 2026. 22 models tracked.

Top models

#ModelScore
1Gemini 3 Pro82.73
2Gemini 2.5 Pro80
3GPT-579.55
4GPT-4o71.36
5Seed 1.867.27
6Qwen 3 VL 32B60.45
7InternVL3.5-8B35

Interactive version: theaggregate.ai/benchmark?slug=gst-bench-top-down-selection-medium · How It Works · Data refreshed daily, snapshot 2026-09-29.