Frontier Models Struggle to Label Robot Egocentric Videos

Tests by Concept and tkkong with MeckaAI found that frontier models struggle to caption egocentric robot videos, with GPT-6 Astra achieving only 43.5% full accuracy; specialized models outperformed frontier ones on the task.

2026-10-13 ~ 2026-10-13 · 2 related posts

• Frontier models can't yet label robot training data: best model only 43.5% fully correct — kimberlywtan · 2026-10-13
• Specialized model beats GPT-6 Astra at captioning egocentric robot videos, best frontier model only 43.5% correct — kimberlywtan · 2026-10-13