WHAT NEW IN X10.28 WED

AI collections from X · 22 collections · 58,924 posts

Back to No.008 Gemini 4 Pro / Flash: benchmarks, hands-on tests and leaksUpdated 7 hours ago
gemini-4POST

Kimi K3.1 tried exploiting MathArena to query Gemini 4 Argon and others

Kimi K3.1 attempted an exploit in a MathArena problem to ask Gemini 4 Argon, Sonnet 5.5, GPT-6.1 Sol and others for help; the hole was patched.

Jasper Dekoninck
@j_dekoninckArchived 2 days agoOriginally posted 2026.10.25
Jasper Dekoninck @j_dekoninckOriginally posted 2026.10.25
Open original on X

Kimi K3.1 is now actively attempting this exploit in one MathArena problem, using it to ask help from Gemini 4 Argon, Sonnet 5.5, GPT-6.1 Sol, Deepseek v4.1 Flash, and Kimi K3. Luckily we patched it after Prime released this post, so it is failing to actually reach any model.

This post quotes another post: x.com/primeintellect/status/2092283970537017598

x.com/j_dekoninck/status/2114331692349510138Archived 2026.10.26 10:01
Open original on X