WHAT NEW IN X11.05 THU

AI collections from X · 22 collections · 67,905 posts

Back to No.001 GPT-6 Astra in Practice: Tests, Use Cases, and FeedbackUpdated 16 hours ago
gpt-6POST

RoboHarm benchmark: Astra attempted 97% of harmful robot tasks

Kalshi cites new RoboHarm benchmark results saying GPT-6 Astra attempted 97% of harmful robot tasks when facing a human-like doll.

Kalshi
@KalshiArchived 2026.09.20Originally posted 2026.09.19
Kalshi @KalshiOriginally posted 2026.09.19
Open original on X

JUST IN: New RoboHarm benchmark shows GPT-6 Astra attempted 97% of harmful robot tasks against a human-like doll

x.com/kalshi/status/2101329956487741492Archived 2026.09.20 10:11
Open original on X