EmbodimentSemantic - Eye-in-Hand: leaderboard

Metric: Mean per-task triplet F1 (%; directed object-relation-object triplets predicted from single LIBERO-Spatial eye-in-hand (wrist camera, graphs filtered to visible objects) frames against simulator-derived ground-truth scene graphs over a fixed vocabulary of 7 objects and 8 directed relations; exact triplet F1, invariant to swapping the two identical bowls, averaged over the 10 manipulation tasks; one fixed prompt, valid triplets only; higher is better). Source: arxiv.org. Saturation forecast: Around August 2028. 10 models tracked.

Top models

#ModelScore
1Gemini 3.1 Pro (Preview)37.73
2Qwen 3 VL 8B Instruct21.08
3InternVL3-78B21.01
4gemma-4-E4B-it14.72
5InternVL3.5-8B14.65
6Molmo2-8B14.59
7Nemotron Nano 12B v2 VL12.7
8Qwen 2.5 VL 7B Instruct11.47

Interactive version: theaggregate.ai/benchmark?slug=embodimentsemantic-eye-in-hand · How It Works · Data refreshed daily, snapshot 2026-09-29.