RoboBrain2.5 8B: benchmark results
Provider: Other.
Aggregate Rating 1588 ± 38, rank #1148 of 10916 rated models, from 27 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| GST-Bench - Egocentric Direction (Visual) | 20.24 | Mean accuracy (%; share of predicted egocentric directions t | 85.7 |
| ERQA-Plus - Navigation and Environmental Reasoning | 78.4 | Multiple-choice accuracy (%) on the 171 navigation and envir | 83.3 |
| GST-Bench - Egocentric Direction (Semantic) | 19.23 | Mean accuracy (%; share of predicted egocentric directions t | 76.2 |
| ERQA-Plus - Social and Interaction Reasoning | 93.3 | Multiple-choice accuracy (%) on the 15 social and interactio | 75 |
| ERQA-Plus | 80.8 | Multiple-choice accuracy (%) over all 1,766 questions in 12 | 66.7 |
| ERQA-Plus - Action-Centric Reasoning | 79 | Multiple-choice accuracy (%) on the 323 action-centric reaso | 66.7 |
| ERQA-Plus - Perceptual Reasoning | 81.7 | Multiple-choice accuracy (%) on the 960 perceptual reasoning | 66.7 |
| GST-Bench - Egocentric Distance (Semantic) | 11.28 | Mean relative accuracy (%; distance to a previously seen obj | 61.9 |
| GST-Bench - Egocentric Distance (Visual) | 10.39 | Mean relative accuracy (%; distance to a previously seen obj | 61.9 |
| GST-Bench | 24.61 | Average score (%; unweighted mean of the 12 subtask scores ( | 52.4 |
| ERQA-Plus - Contextual and Commonsense Reasoning | 80.5 | Multiple-choice accuracy (%) on the 297 contextual and commo | 50 |
| ERQA-Plus - Open-Ended | 55.7 | Mean Sentence-BERT cosine similarity (x100, -100 to 100) bet | 50 |
Interactive version: theaggregate.ai/model?slug=robobrain2-5-8b · How It Works · Data refreshed daily, snapshot 2026-11-08.