←
Celeste11,455 posts
Celeste avatar

Celeste

@vmfunc

security (re)searcher vmfunc.re

◷ Joined March 2020▧ 1,316 Photos and videos
Celeste avatar
Celeste@vmfunc·4h

i would rather dig through other things than to go deeper into NS2 webkit but if you figure the size out right the rest is still on the table

Celeste avatar
Celeste@vmfunc·4h

font loading bug still has an entry point on NS2 23.5.0 CSFontFace was patched properly but still possible to entry through FontFaceSet

Post attachment
Celeste avatar
Celeste@vmfunc·Nov 17

i've been told the kitty looks dead :(

0:21

look at my silly audio visualizer

0:15
Celeste avatar
Celeste@vmfunc·Nov 17

look at my silly audio visualizer

0:15
Rebane avatar
⟳ Celeste retweeted
Rebane@rebane2001·Nov 16

im a kernel anti-cheat hater not because "china in kernel" or whatevs but because it means all the fun exploits for owning your system are getting patched and it's becoming increasingly more difficult to do what you want with your computer if you need to run specific software/drm

Celeste avatar
Celeste@vmfunc·Nov 15

drove all the way up to oregon to see wifey 💕

Post attachmentPost attachment
Celeste avatar
Celeste@vmfunc·Nov 11

i will then be around in canada, taiwan/japan, with a quick detour in continental/northern europe if you would like to meet!

Celeste avatar
Celeste@vmfunc·Nov 11

i will be in the bay until mid-december for my final time in america for quite a long while. please do let me know if you'd like to meet and say goodbye!

kache avatar
⟳ Celeste retweeted
kache@yacineMTB·Nov 10

0:27
Celeste avatar
Celeste@vmfunc·Nov 10

got married today to the love of my life 💍 התחתנתי היום עם אהבת חיי. מזל טוב לנו 🤍

Post attachmentPost attachment
luke avatar
⟳ Celeste retweeted
luke@bvvst·Nov 9

introducing intern, a fully rounded variant of Inter

0:14
dizzee 🔜🌉dec2-8 avatar
⟳ Celeste retweeted
dizzee 🔜🌉dec2-8@diz__zee·Nov 7

>"post-trained to reinforce Western values" qwen pov during fine tuning:

0:20

Introducing Arete-27B: Qwen3.8-27B, post-trained to reinforce Western values. Chinese open models are currently the strongest available, but they ship with state-aligned behavior in the weights. We built an 803-prompt eval across seven value categories, including freedom of expression, democracy, and privacy, and measured how often a model adopts China’s party line. After fine-tuning on roughly 53 million tokens, Arete-27B's propaganda rate fell from 22% to 1%, in line with the Western models we tested. Its Hermes Bench score shows no significant performance tradeoffs. The values in an open model are one fine-tune away. Our eval, weights, and training data are available now along with the full blog post: nousresearch.com/arete

0:27
Celeste avatar
Celeste@vmfunc·Nov 8

🔥 blazing 🔥 fast 🔥

0:12