One VLA policy places 9th in the 2026 BEHAVIOR Challenge
A single π0.5-based checkpoint scores 0.392 Q on 100 tasks; GPT-6 Astra averages 0.639 Q on 14 routed tasks.
Stories per day · 30 days
1 stories, newest first · one card per story
1 stories, newest first · one card per story
A single π0.5-based checkpoint scores 0.392 Q on 100 tasks; GPT-6 Astra averages 0.639 Q on 14 routed tasks.
Today 707 readers · 836 views · 275 reads · all time 187K views